Artificial Intelligence Nanodegree

Computer Vision Capstone

Project: Facial Keypoint Detection

Welcome to the final Computer Vision project in the Artificial Intelligence Nanodegree program!

In this project, you’ll combine your knowledge of computer vision techniques and deep learning to build and end-to-end facial keypoint recognition system! Facial keypoints include points around the eyes, nose, and mouth on any face and are used in many applications, from facial tracking to emotion recognition.

There are three main parts to this project:

Part 1 : Investigating OpenCV, pre-processing, and face detection

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!

*Here's what you need to know to complete the project:

  1. In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested.

    a. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

  1. In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation.

    a. Each section where you will answer a question is preceded by a 'Question X' header.

    b. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional suggestions for enhancing the project beyond the minimum requirements. If you decide to pursue the "(Optional)" sections, you should include the code in this IPython notebook.

Your project submission will be evaluated based on your answers to each of the questions and the code implementations you provide.

Steps to Complete the Project

Each part of the notebook is further broken down into separate steps. Feel free to use the links below to navigate the notebook.

In this project you will get to explore a few of the many computer vision algorithms built into the OpenCV library. This expansive computer vision library is now almost 20 years old and still growing!

The project itself is broken down into three large parts, then even further into separate steps. Make sure to read through each step, and complete any sections that begin with '(IMPLEMENTATION)' in the header; these implementation sections may contain multiple TODOs that will be marked in code. For convenience, we provide links to each of these steps below.

Part 1 : Investigating OpenCV, pre-processing, and face detection

  • Step 0: Detect Faces Using a Haar Cascade Classifier
  • Step 1: Add Eye Detection
  • Step 2: De-noise an Image for Better Face Detection
  • Step 3: Blur an Image and Perform Edge Detection
  • Step 4: Automatically Hide the Identity of an Individual

Part 2 : Training a Convolutional Neural Network (CNN) to detect facial keypoints

  • Step 5: Create a CNN to Recognize Facial Keypoints
  • Step 6: Compile and Train the Model
  • Step 7: Visualize the Loss and Answer Questions

Part 3 : Putting parts 1 and 2 together to identify facial keypoints on any image!

  • Step 8: Build a Robust Facial Keypoints Detector (Complete the CV Pipeline)

Step 0: Detect Faces Using a Haar Cascade Classifier

Have you ever wondered how Facebook automatically tags images with your friends' faces? Or how high-end cameras automatically find and focus on a certain person's face? Applications like these depend heavily on the machine learning task known as face detection - which is the task of automatically finding faces in images containing people.

At its root face detection is a classification problem - that is a problem of distinguishing between distinct classes of things. With face detection these distinct classes are 1) images of human faces and 2) everything else.

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the detector_architectures directory.

Import Resources

In the next python cell, we load in the required libraries for this section of the project.

In [36]:
# Import required libraries for this section

%matplotlib inline

import numpy as np
import matplotlib.pyplot as plt
import math
import cv2                     # OpenCV library for computer vision
from PIL import Image
import time

Next, we load in and display a test image for performing face detection.

Note: by default OpenCV assumes the ordering of our image's color channels are Blue, then Green, then Red. This is slightly out of order with most image types we'll use in these experiments, whose color channels are ordered Red, then Green, then Blue. In order to switch the Blue and Red channels of our test image around we will use OpenCV's cvtColor function, which you can read more about by checking out some of its documentation located here. This is a general utility function that can do other transformations too like converting a color image to grayscale, and transforming a standard color image to HSV color space.

In [37]:
# Load in color image for face detection
image = cv2.imread('images/test_image_1.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot our image using subplots to specify a size and title
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)
Out[37]:
<matplotlib.image.AxesImage at 0x7fd68305cda0>

There are a lot of people - and faces - in this picture. 13 faces to be exact! In the next code cell, we demonstrate how to use a Haar Cascade classifier to detect all the faces in this test image.

This face detector uses information about patterns of intensity in an image to reliably detect faces under varying light conditions. So, to use this face detector, we'll first convert the image from color to grayscale.

Then, we load in the fully trained architecture of the face detector -- found in the file haarcascade_frontalface_default.xml - and use it on our image to find faces!

To learn more about the parameters of the detector see this post.

In [38]:
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detections')
ax1.imshow(image_with_detections)

Number of faces detected: 13
Out[38]:
<matplotlib.image.AxesImage at 0x7fd68300d080>

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Step 1: Add Eye Detections

There are other pre-trained detectors available that use a Haar Cascade Classifier - including full human body detectors, license plate detectors, and more. A full list of the pre-trained architectures can be found here.

To test your eye detector, we'll first read in a new test image with just a single face.

In [39]:
# Load in color image for face detection
image = cv2.imread('images/james.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Plot the RGB image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)
Out[39]:
<matplotlib.image.AxesImage at 0x7fd6830332e8>

Notice that even though the image is a black and white image, we have read it in as a color image and so it will still need to be converted to grayscale in order to perform the most accurate face detection.

So, the next steps will be to convert this image to grayscale, then load OpenCV's face detector and run it with parameters that detect this face accurately.

In [40]:
# Convert the RGB  image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray, 1.25, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face Detection')
ax1.imshow(image_with_detections)

Number of faces detected: 1
Out[40]:
<matplotlib.image.AxesImage at 0x7fd682fde358>

(IMPLEMENTATION) Add an eye detector to the current face detection setup.

A Haar-cascade eye detector can be included in the same way that the face detector was and, in this first task, it will be your job to do just this.

To set up an eye detector, use the stored parameters of the eye cascade detector, called haarcascade_eye.xml, located in the detector_architectures subdirectory. In the next code cell, create your eye detector and store its detections.

A few notes before you get started:

First, make sure to give your loaded eye detector the variable name

eye_cascade

and give the list of eye regions you detect the variable name

eyes

Second, since we've already run the face detector over this image, you should only search for eyes within the rectangular face regions detected in faces. This will minimize false detections.

Lastly, once you've run your eye detector over the facial detection region, you should display the RGB image with both the face detection boxes (in red) and your eye detections (in green) to verify that everything works as expected.

In [41]:
# Make a copy of the original image to plot rectangle detections
image_with_detections = np.copy(image)   

# Loop over the detections and draw their corresponding face detection boxes
for (x,y,w,h) in faces:
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h),(255,0,0), 3)  
    
# Do not change the code above this comment!
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

## TODO: Add eye detection, using haarcascade_eye.xml, to the current face detector algorithm
eye_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_eye.xml')


## TODO: Loop over the eye detections and draw their corresponding boxes in green on image_with_detections
eyes = eye_cascade.detectMultiScale(gray, 1.1, 6)

for (x,y,w,h) in eyes:
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h),(255,0,0), 1)  


# Plot the image with both faces and eyes detected
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Image with Face and Eye Detection')
ax1.imshow(image_with_detections)
Out[41]:
<matplotlib.image.AxesImage at 0x7fd682f84f98>

(Optional) Add face and eye detection to your laptop camera

It's time to kick it up a notch, and add face and eye detection to your laptop's camera! Afterwards, you'll be able to show off your creation like in the gif shown below - made with a completed version of the code!

Notice that not all of the detections here are perfect - and your result need not be perfect either. You should spend a small amount of time tuning the parameters of your detectors to get reasonable results, but don't hold out for perfection. If we wanted perfection we'd need to spend a ton of time tuning the parameters of each detector, cleaning up the input image frames, etc. You can think of this as more of a rapid prototype.

The next cell contains code for a wrapper function called laptop_camera_face_eye_detector that, when called, will activate your laptop's camera. You will place the relevant face and eye detection code in this wrapper function to implement face/eye detection and mark those detections on each image frame that your camera captures.

Before adding anything to the function, you can run it to get an idea of how it works - a small window should pop up showing you the live feed from your camera; you can press any key to close this window.

Note: Mac users may find that activating this function kills the kernel of their notebook every once in a while. If this happens to you, just restart your notebook's kernel, activate cell(s) containing any crucial import statements, and you'll be good to go!

In [42]:
def face_eye_detector(image):
    img = np.copy(image)
    rgb_image = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    gray = cv2.cvtColor(rgb_image,cv2.COLOR_RGB2GRAY)
    face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')
    faces = face_cascade.detectMultiScale(gray, 1.25, 6)
    # Print the number of faces detected in the image
    #print('Number of faces detected:', len(faces))
    # Add eye detection, using haarcascade_eye.xml, to the current face detector algorithm
    eye_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_eye.xml')
    eyes = eye_cascade.detectMultiScale(gray, 1.1, 6)
   
    # Make a copy of the orginal image to draw face detections on
    image_with_detections = np.copy(image)

    # Get the bounding box for each detected face & eyes
    for (x,y,w,h) in faces:
  
        cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)  
        eyes = eye_cascade.detectMultiScale(gray[y:y+h,x:x+w], 1.1, 6)
        for (xe,ye,we,he) in eyes:
            cv2.rectangle(image_with_detections[y:y+h,x:x+w], (xe,ye), (xe+we,ye+he),(255,0,0), 1)
            
    return image_with_detections
In [43]:
### Add face and eye detection to this laptop camera function 
# Make sure to draw out all faces/eyes found in each frame on the shown video feed

import cv2
### Insert face detection and blurring code into the wrapper below to create an identity protector on your laptop!
import cv2
import time 
import matplotlib.pyplot as plt

def laptop_camera_go_face_eyes():
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    while rval:
        frame = face_eye_detector(frame)
        cv2.imshow("face detection activated", frame)
        key = cv2.waitKey(20)
        if key == 27: # esc 
            cv2.destroyAllWindows()
            for i in range (1,5):
                cv2.waitKey(1)
            return
        time.sleep(0.01)            
        rval, frame = vc.read()
In [44]:
# Call the laptop camera face/eye detector function above
laptop_camera_go()

Face with Eyes

Step 2: De-noise an Image for Better Face Detection

Image quality is an important aspect of any computer vision task. Typically, when creating a set of images to train a deep learning network, significant care is taken to ensure that training images are free of visual noise or artifacts that hinder object detection. While computer vision algorithms - like a face detector - are typically trained on 'nice' data such as this, new test data doesn't always look so nice!

When applying a trained computer vision algorithm to a new piece of test data one often cleans it up first before feeding it in. This sort of cleaning - referred to as pre-processing - can include a number of cleaning phases like blurring, de-noising, color transformations, etc., and many of these tasks can be accomplished using OpenCV.

In this short subsection we explore OpenCV's noise-removal functionality to see how we can clean up a noisy image, which we then feed into our trained face detector.

Create a noisy image to work with

In the next cell, we create an artificial noisy version of the previous multi-face image. This is a little exaggerated - we don't typically get images that are this noisy - but image noise, or 'grainy-ness' in a digitial image - is a fairly common phenomenon.

In [45]:
# Load in the multi-face test image again
image = cv2.imread('images/test_image_1.jpg')

# Convert the image copy to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Make an array copy of this image
image_with_noise = np.asarray(image)

# Create noise - here we add noise sampled randomly from a Gaussian distribution: a common model for noise
noise_level = 40
noise = np.random.randn(image.shape[0],image.shape[1],image.shape[2])*noise_level

# Add this noise to the array image copy
image_with_noise = image_with_noise + noise

# Convert back to uint8 format
image_with_noise = np.asarray([np.uint8(np.clip(i,0,255)) for i in image_with_noise])

# Plot our noisy image!
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image')
ax1.imshow(image_with_noise)
Out[45]:
<matplotlib.image.AxesImage at 0x7fd682f39978>

In the context of face detection, the problem with an image like this is that - due to noise - we may miss some faces or get false detections.

In the next cell we apply the same trained OpenCV detector with the same settings as before, to see what sort of detections we get.

In [46]:
def display_images(num = None,image1=None,title1=None,cmap1=None,image2=None,title2=None,cmap2=None):
    if num == 2:
        fig = plt.figure(figsize = (15,15))
        ax1 = fig.add_subplot(121)
        ax1.set_xticks([])
        ax1.set_yticks([])

        ax1.set_title(title1)
        ax1.imshow(image1,cmap = cmap1)

        ax2 = fig.add_subplot(122)
        ax2.set_xticks([])
        ax2.set_yticks([])

        ax2.set_title(title2)
        ax2.imshow(image2, cmap=cmap2)
    elif num == 1:
        fig = plt.figure(figsize = (8,8))
        ax1 = fig.add_subplot(121)
        ax1.set_xticks([])
        ax1.set_yticks([])

        ax1.set_title(title1)
        ax1.imshow(image1,cmap = cmap1)
In [47]:
# Convert the RGB  image to grayscale
gray_noise = cv2.cvtColor(image_with_noise, cv2.COLOR_RGB2GRAY)

# Extract the pre-trained face detector from an xml file
face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

# Detect the faces in image
faces = face_cascade.detectMultiScale(gray_noise, 4, 6)

# Print the number of faces detected in the image
print('Number of faces detected:', len(faces))

# Make a copy of the orginal image to draw face detections on
image_with_detections = np.copy(image_with_noise)

# Get the bounding box for each detected face
for (x,y,w,h) in faces:
    # Add a red bounding box to the detections image
    cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)
    

# Display the image with the detections
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Noisy Image with Face Detections')
ax1.imshow(image_with_detections)

Number of faces detected: 11
Out[47]:
<matplotlib.image.AxesImage at 0x7fd682f66128>

With this added noise we now miss one of the faces!

(IMPLEMENTATION) De-noise this image for better face detection

Time to get your hands dirty: using OpenCV's built in color image de-noising functionality called fastNlMeansDenoisingColored - de-noise this image enough so that all the faces in the image are properly detected. Once you have cleaned the image in the next cell, use the cell that follows to run our trained face detector over the cleaned image to check out its detections.

You can find its official documentation here and a useful example here.

Note: you can keep all parameters except photo_render fixed as shown in the second link above. Play around with the value of this parameter - see how it affects the resulting cleaned image.

In [48]:
## TODO: Use OpenCV's built in color image de-noising function to clean up our noisy image!
denoised_image = cv2.fastNlMeansDenoisingColored(image_with_noise,None,15,10,7,30)
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.imshow(denoised_image)
Out[48]:
<matplotlib.image.AxesImage at 0x7fd6c7fd2c88>
In [49]:
## TODO: Run the face detector on the de-noised image to improve your detections and display the result
def face_detect_and_draw(image):
    ## TODO: Run the face detector on the de-noised image to improve your detections and display the result
    # Convert the RGB  image to grayscale
    gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

    # Extract the pre-trained face detector from an xml file
    face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

    # Detect the faces in image
    faces = face_cascade.detectMultiScale(gray, 4, 6)

    # Print the number of faces detected in the image
    print('Number of faces detected:', len(faces))

    # Make a copy of the orginal image to draw face detections on
    image_with_detections = np.copy(image)

    # Get the bounding box for each detected face
    for (x,y,w,h) in faces:
        # Add a red bounding box to the detections image
        cv2.rectangle(image_with_detections, (x,y), (x+w,y+h), (255,0,0), 3)


    # Display the image with the detections
    fig = plt.figure(figsize = (8,8))
    ax1 = fig.add_subplot(111)
    ax1.set_xticks([])
    ax1.set_yticks([])

    ax1.set_title('Image with Face Detections')
    ax1.imshow(image_with_detections)
In [50]:
face_detect_and_draw(denoised_image)

Number of faces detected: 13

Step 3: Blur an Image and Perform Edge Detection

Now that we have developed a simple pipeline for detecting faces using OpenCV - let's start playing around with a few fun things we can do with all those detected faces!

Importance of Blur in Edge Detection

Edge detection is a concept that pops up almost everywhere in computer vision applications, as edge-based features (as well as features built on top of edges) are often some of the best features for e.g., object detection and recognition problems.

Edge detection is a dimension reduction technique - by keeping only the edges of an image we get to throw away a lot of non-discriminating information. And typically the most useful kind of edge-detection is one that preserves only the important, global structures (ignoring local structures that aren't very discriminative). So removing local structures / retaining global structures is a crucial pre-processing step to performing edge detection in an image, and blurring can do just that.

Below is an animated gif showing the result of an edge-detected cat taken from Wikipedia, where the image is gradually blurred more and more prior to edge detection. When the animation begins you can't quite make out what it's a picture of, but as the animation evolves and local structures are removed via blurring the cat becomes visible in the edge-detected image.

Edge detection is a convolution performed on the image itself, and you can read about Canny edge detection on this OpenCV documentation page.

Canny edge detection

In the cell below we load in a test image, then apply Canny edge detection on it. The original image is shown on the left panel of the figure, while the edge-detected version of the image is shown on the right. Notice how the result looks very busy - there are too many little details preserved in the image before it is sent to the edge detector. When applied in computer vision applications, edge detection should preserve global structure; doing away with local structures that don't help describe what objects are in the image.

In [51]:
# Load in the image
image = cv2.imread('images/fawzia.jpg')

# Convert to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)  

# Perform Canny edge detection
edges = cv2.Canny(gray,100,200)

# Dilate the image to amplify edges
edges = cv2.dilate(edges, None)

# Plot the RGB and edge-detected image
fig = plt.figure(figsize = (15,15))
ax1 = fig.add_subplot(121)
ax1.set_xticks([])
ax1.set_yticks([])

ax1.set_title('Original Image')
ax1.imshow(image)

ax2 = fig.add_subplot(122)
ax2.set_xticks([])
ax2.set_yticks([])

ax2.set_title('Canny Edges')
ax2.imshow(edges, cmap='gray')
Out[51]:
<matplotlib.image.AxesImage at 0x7fd6c7f46f60>

Without first blurring the image, and removing small, local structures, a lot of irrelevant edge content gets picked up and amplified by the detector (as shown in the right panel above).

(IMPLEMENTATION) Blur the image then perform edge detection

In the next cell, you will repeat this experiment - blurring the image first to remove these local structures, so that only the important boudnary details remain in the edge-detected image.

Blur the image by using OpenCV's filter2d functionality - which is discussed in this documentation page - and use an averaging kernel of width equal to 4.

In [52]:
### TODO: Blur the test imageusing OpenCV's filter2d functionality, 
# Use an averaging kernel, and a kernel width equal to 4

kernel = np.ones((4,4),np.float32)/25
blurred = cv2.filter2D(image,-1,kernel)
     
display_images(2,image,'Original',None,blurred,'Blurred',None)
## TODO: Then perform Canny edge detection and display the output
upper = 120
lower = 60
cannyed = cv2.Canny(blurred,lower,upper)
display_images(2,blurred,'Blurred',None,cannyed,'Canny',None)

Step 4: Automatically Hide the Identity of an Individual

If you film something like a documentary or reality TV, you must get permission from every individual shown on film before you can show their face, otherwise you need to blur it out - by blurring the face a lot (so much so that even the global structures are obscured)! This is also true for projects like Google's StreetView maps - an enormous collection of mapping images taken from a fleet of Google vehicles. Because it would be impossible for Google to get the permission of every single person accidentally captured in one of these images they blur out everyone's faces, the detected images must automatically blur the identity of detected people. Here's a few examples of folks caught in the camera of a Google street view vehicle.

Read in an image to perform identity detection

Let's try this out for ourselves. Use the face detection pipeline built above and what you know about using the filter2D to blur and image, and use these in tandem to hide the identity of the person in the following image - loaded in and printed in the next cell.

In [53]:
# Load in the image
image = cv2.imread('images/gus.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Display the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)
Out[53]:
<matplotlib.image.AxesImage at 0x7fd6c7f66198>

(IMPLEMENTATION) Use blurring to hide the identity of an individual in an image

The idea here is to 1) automatically detect the face in this image, and then 2) blur it out! Make sure to adjust the parameters of the averaging blur filter to completely obscure this person's identity.

In [54]:
## TODO: Implement face detection
def face_detector(image,if_gray=False,face_detection_val = (1.26,6)):
    ## TODO: Run the face detector on the de-noised image to improve your detections and display the result
    # Convert the RGB  image to grayscale
    if if_gray == True:
        gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
    else:
        gray = np.copy(image)
    # Extract the pre-trained face detector from an xml file
    face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

    # Detect the faces in image
    
    faces = face_cascade.detectMultiScale(gray, face_detection_val[0], face_detection_val[1])

    # Print the number of faces detected in the image
    print('Number of faces detected:', len(faces))
    return faces

def draw_faces(image,faces):
    image_with_faces = np.copy(image)

    # Get the bounding box for each detected face
    for (x,y,w,h) in faces:
        # Add a red bounding box to the detections image
        cv2.rectangle(image_with_faces, (x,y), (x+w,y+h), (255,0,0), 3)
        print(x,y,w,h)
    return image_with_faces

## TODO: Blur the bounding box around each detected face using an averaging filter and display the result
def blur_single_small_area_2dconv(image=None,blurred_area=None, kernel_edge=None):
    img = np.copy(image)
    print(img.shape)
    kernel = np.ones((kernel_edge,kernel_edge),np.float32)/(kernel_edge*kernel_edge)
    blurred = None
    if blurred_area is None:
        blurred = cv2.filter2D(img,-1,kernel)
    else:
        x,y,w,h = blurred_area
        print(x,y,w,h)
        blurred = cv2.filter2D(img[y:y+h,x:x+w],-1,kernel)
        #display_images(2,img,'Image',None,img[y:y+h,x:x+w],'The face',None)
    return blurred 


def face_blurrer(image,intensity):
    img = np.copy(image)
    faces = face_detector(img)
    for (x,y,w,h) in faces:
        blurred = blur_single_small_area_2dconv(img,(x,y,w,h),intensity)
        img[y:y+h,x:x+w] = blurred
       
    return img
In [55]:
faces = face_detector(image,False,(4,6))
image_with_faces = draw_faces(image,faces)
display_images(1,image_with_faces,'Face Detected',None)

Number of faces detected: 1
773 103 384 384
In [56]:
## TODO: Blur the bounding box around each detected face using an averaging filter and display the result
img = face_blurrer(image,200)
display_images(1,img,'Blurred Face',None)

Number of faces detected: 1
(1334, 2000, 3)
778 108 374 374

(Optional) Build identity protection into your laptop camera

In this optional task you can add identity protection to your laptop camera, using the previously completed code where you added face detection to your laptop camera - and the task above. You should be able to get reasonable results with little parameter tuning - like the one shown in the gif below.

As with the previous video task, to make this perfect would require significant effort - so don't strive for perfection here, strive for reasonable quality.

The next cell contains code a wrapper function called laptop_camera_identity_hider that - when called - will activate your laptop's camera. You need to place the relevant face detection and blurring code developed above in this function in order to blur faces entering your laptop camera's field of view.

Before adding anything to the function you can call it to get a hang of how it works - a small window will pop up showing you the live feed from your camera, you can press any key to close this window.

Note: Mac users may find that activating this function kills the kernel of their notebook every once in a while. If this happens to you, just restart your notebook's kernel, activate cell(s) containing any crucial import statements, and you'll be good to go!

In [57]:
import numpy as np
import cv2
cams_test = 10
for i in range(0,cams_test):
    cap = cv2.VideoCapture(i)
    test,frame = cap.read()
    print("i : " + str(i) + "////result:" + str(test))

i : 0////result:False
i : 1////result:False
i : 2////result:False
i : 3////result:False
i : 4////result:False
i : 5////result:False
i : 6////result:False
i : 7////result:False
i : 8////result:False
i : 9////result:False
In [58]:
### Insert face detection and blurring code into the wrapper below to create an identity protector on your laptop!
import cv2
import time 
import matplotlib.pyplot as plt

def laptop_camera_go():
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    while rval:
        frame = face_blurrer(frame,200)
        cv2.imshow("face detection activated", frame)
        key = cv2.waitKey(20)
        if key == 27 : # esc
            cv2.destroyAllWindows()
           
            for i in range (1,5):
                cv2.waitKey(1)
            return
        time.sleep(0.01)             
        rval, frame = vc.read()
In [59]:
# Run laptop identity hider
laptop_camera_go()

Face with Sunglasses

Step 5: Create a CNN to Recognize Facial Keypoints

OpenCV is often used in practice with other machine learning and deep learning libraries to produce interesting results. In this stage of the project you will create your own end-to-end pipeline - employing convolutional networks in keras along with OpenCV - to apply a "selfie" filter to streaming video and images.

You will start by creating and then training a convolutional network that can detect facial keypoints in a small dataset of cropped images of human faces. We then guide you towards OpenCV to expanding your detection algorithm to more general images. What are facial keypoints? Let's take a look at some examples.

Facial keypoints (also called facial landmarks) are the small blue-green dots shown on each of the faces in the image above - there are 15 keypoints marked in each image. They mark important areas of the face - the eyes, corners of the mouth, the nose, etc. Facial keypoints can be used in a variety of machine learning applications from face and emotion recognition to commercial applications like the image filters popularized by Snapchat.

Below we illustrate a filter that, using the results of this section, automatically places sunglasses on people in images (using the facial keypoints to place the glasses correctly on each face). Here, the facial keypoints have been colored lime green for visualization purposes.

Make a facial keypoint detector

But first things first: how can we make a facial keypoint detector? Well, at a high level, notice that facial keypoint detection is a regression problem. A single face corresponds to a set of 15 facial keypoints (a set of 15 corresponding $(x, y)$ coordinates, i.e., an output point). Because our input data are images, we can employ a convolutional neural network to recognize patterns in our images and learn how to identify these keypoint given sets of labeled data.

In order to train a regressor, we need a training set - a set of facial image / facial keypoint pairs to train on. For this we will be using this dataset from Kaggle. We've already downloaded this data and placed it in the data directory. Make sure that you have both the training and test data files. The training dataset contains several thousand $96 \times 96$ grayscale images of cropped human faces, along with each face's 15 corresponding facial keypoints (also called landmarks) that have been placed by hand, and recorded in $(x, y)$ coordinates. This wonderful resource also has a substantial testing set, which we will use in tinkering with our convolutional network.

To load in this data, run the Python cell below - notice we will load in both the training and testing sets.

The load_data function is in the included utils.py file.

In [60]:
import keras
import matplotlib as plt
In [61]:
from keras.utils import np_utils
from utils import *
In [62]:
# Load training set
X_train, y_train = load_data()
print("X_train.shape == {}".format(X_train.shape))
print("y_train.shape == {}; y_train.min == {:.3f}; y_train.max == {:.3f}".format(
    y_train.shape, y_train.min(), y_train.max()))

# Load testing set
X_test,_ = load_data(test=True)
print("X_test.shape == {}".format(X_test.shape))

X_train.shape == (2140, 96, 96, 1)
y_train.shape == (2140, 30); y_train.min == -0.920; y_train.max == 0.996
X_test.shape == (1783, 96, 96, 1)

The load_data function in utils.py originates from this excellent blog post, which you are strongly encouraged to read. Please take the time now to review this function. Note how the output values - that is, the coordinates of each set of facial landmarks - have been normalized to take on values in the range $[-1, 1]$, while the pixel values of each input point (a facial image) have been normalized to the range $[0,1]$.

Note: the original Kaggle dataset contains some images with several missing keypoints. For simplicity, the load_data function removes those images with missing labels from the dataset. As an optional extension, you are welcome to amend the load_data function to include the incomplete data points.

Visualize the Training Data

Execute the code cell below to visualize a subset of the training data.

In [63]:
import matplotlib.pyplot as plt
%matplotlib inline

fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_train[i], y_train[i], ax)

For each training image, there are two landmarks per eyebrow (four total), three per eye (six total), four for the mouth, and one for the tip of the nose.

Review the plot_data function in utils.py to understand how the 30-dimensional training labels in y_train are mapped to facial locations, as this function will prove useful for your pipeline.

(IMPLEMENTATION) Specify the CNN Architecture

In this section, you will specify a neural network for predicting the locations of facial keypoints. Use the code cell below to specify the architecture of your neural network. We have imported some layers that you may find useful for this task, but if you need to use more Keras layers, feel free to import them in the cell.

Your network should accept a $96 \times 96$ grayscale image as input, and it should output a vector with 30 entries, corresponding to the predicted (horizontal and vertical) locations of 15 facial keypoints. If you are not sure where to start, you can find some useful starting architectures in this blog, but you are not permitted to copy any of the architectures that you find online.

In [64]:
# Import deep learning resources from Keras
from keras.models import Sequential
from keras.layers import Convolution2D, MaxPooling2D, Dropout
from keras.layers import Flatten, Dense


## TODO: Specify a CNN architecture
# Your model should accept 96x96 pixel graysale images in
# It should have a fully-connected output layer with 30 values (2 for each facial keypoint)
model = Sequential()
model.add(Convolution2D(filters=32, kernel_size=3, padding='same', activation='relu', input_shape=(96, 96, 1)))
model.add(MaxPooling2D(pool_size=2))
model.add(Convolution2D(filters=16, kernel_size=3, padding='same', activation='relu', input_shape=(96, 96, 1)))
model.add(MaxPooling2D(pool_size=2))
model.add(Flatten())
model.add(Dropout(0.5))
model.add(Dense(30))
# Summarize the model
model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_3 (Conv2D)            (None, 96, 96, 32)        320       
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 48, 48, 32)        0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 48, 48, 16)        4624      
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 24, 24, 16)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 9216)              0         
_________________________________________________________________
dropout_2 (Dropout)          (None, 9216)              0         
_________________________________________________________________
dense_2 (Dense)              (None, 30)                276510    
=================================================================
Total params: 281,454
Trainable params: 281,454
Non-trainable params: 0
_________________________________________________________________

Step 6: Compile and Train the Model

After specifying your architecture, you'll need to compile and train the model to detect facial keypoints'

(IMPLEMENTATION) Compile and Train the Model

Use the compile method to configure the learning process. Experiment with your choice of optimizer; you may have some ideas about which will work best (SGD vs. RMSprop, etc), but take the time to empirically verify your theories.

Use the fit method to train the model. Break off a validation set by setting validation_split=0.2. Save the returned History object in the history variable.

Experiment with your model to minimize the validation loss (measured as mean squared error). A very good model will achieve about 0.0015 loss (though it's possible to do even better). When you have finished training, save your model as an HDF5 file with file path my_model.h5.

In [65]:
from keras.optimizers import SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam
from keras.callbacks import ModelCheckpoint ,EarlyStopping

opti =['rmsprop','adam','sgd','adagrad','adadelta','adamax']
hist = {}
epoc = 100
In [10]:
## TODO: Train the model  #reduce learni
for opt in opti:
    checkpointer = ModelCheckpoint(filepath='my_model.{}.h5'.format(opt), verbose=1, save_best_only=True)
    earlystopping = EarlyStopping(monitor='val_loss',min_delta=0,patience=30,verbose=1, mode='auto')
    model.compile(optimizer= opt, loss='mean_squared_error', metrics=['accuracy'] )
    hist[opt] = model.fit(X_train, y_train, batch_size=128, epochs=epoc,validation_split =0.2, callbacks=[checkpointer,earlystopping], verbose=1, shuffle=True)

Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0680 - acc: 0.5036Epoch 00000: val_loss improved from inf to 0.01621, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 2s - loss: 0.0666 - acc: 0.5018 - val_loss: 0.0162 - val_acc: 0.6963
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0192 - acc: 0.6226Epoch 00001: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0192 - acc: 0.6250 - val_loss: 0.0208 - val_acc: 0.6963
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0201 - acc: 0.6412Epoch 00002: val_loss improved from 0.01621 to 0.01316, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0199 - acc: 0.6414 - val_loss: 0.0132 - val_acc: 0.6963
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0172 - acc: 0.6641Epoch 00003: val_loss improved from 0.01316 to 0.00965, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0171 - acc: 0.6659 - val_loss: 0.0097 - val_acc: 0.6939
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0145 - acc: 0.6653Epoch 00004: val_loss improved from 0.00965 to 0.00847, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0146 - acc: 0.6659 - val_loss: 0.0085 - val_acc: 0.6963
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0114 - acc: 0.6851Epoch 00005: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0115 - acc: 0.6863 - val_loss: 0.0096 - val_acc: 0.6916
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0108 - acc: 0.6815Epoch 00006: val_loss improved from 0.00847 to 0.00512, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0106 - acc: 0.6822 - val_loss: 0.0051 - val_acc: 0.7033
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0106 - acc: 0.6965Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0106 - acc: 0.6933 - val_loss: 0.0055 - val_acc: 0.6916
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0080 - acc: 0.6965Epoch 00008: val_loss improved from 0.00512 to 0.00408, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0080 - acc: 0.6968 - val_loss: 0.0041 - val_acc: 0.6963
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0075 - acc: 0.7133Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0076 - acc: 0.7138 - val_loss: 0.0055 - val_acc: 0.6963
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0068 - acc: 0.7109Epoch 00010: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0069 - acc: 0.7114 - val_loss: 0.0064 - val_acc: 0.7150
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0065 - acc: 0.7133Epoch 00011: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0065 - acc: 0.7109 - val_loss: 0.0049 - val_acc: 0.7056
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0061 - acc: 0.7151Epoch 00012: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0061 - acc: 0.7144 - val_loss: 0.0050 - val_acc: 0.6986
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0055 - acc: 0.7218Epoch 00013: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0056 - acc: 0.7196 - val_loss: 0.0055 - val_acc: 0.7220
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0049 - acc: 0.7169Epoch 00014: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0049 - acc: 0.7179 - val_loss: 0.0048 - val_acc: 0.7383
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0052 - acc: 0.7266Epoch 00015: val_loss improved from 0.00408 to 0.00326, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0053 - acc: 0.7255 - val_loss: 0.0033 - val_acc: 0.7266
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0044 - acc: 0.7163Epoch 00016: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0045 - acc: 0.7144 - val_loss: 0.0035 - val_acc: 0.7243
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0043 - acc: 0.7266Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0044 - acc: 0.7284 - val_loss: 0.0041 - val_acc: 0.7196
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0041 - acc: 0.7356Epoch 00018: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0041 - acc: 0.7348 - val_loss: 0.0037 - val_acc: 0.7126
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0039 - acc: 0.7296Epoch 00019: val_loss improved from 0.00326 to 0.00320, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0039 - acc: 0.7266 - val_loss: 0.0032 - val_acc: 0.7173
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0037 - acc: 0.7386Epoch 00020: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0037 - acc: 0.7389 - val_loss: 0.0033 - val_acc: 0.7313
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0048 - acc: 0.7338Epoch 00021: val_loss improved from 0.00320 to 0.00261, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0048 - acc: 0.7360 - val_loss: 0.0026 - val_acc: 0.7220
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0022 - acc: 0.7284Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7284 - val_loss: 0.0072 - val_acc: 0.7360
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0036 - acc: 0.7404Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7389 - val_loss: 0.0030 - val_acc: 0.7313
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0031 - acc: 0.7482Epoch 00024: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0032 - acc: 0.7488 - val_loss: 0.0032 - val_acc: 0.7290
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0031 - acc: 0.7464Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0031 - acc: 0.7465 - val_loss: 0.0030 - val_acc: 0.7336
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0034 - acc: 0.7506Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0034 - acc: 0.7518 - val_loss: 0.0028 - val_acc: 0.7290
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0029 - acc: 0.7560Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7558 - val_loss: 0.0033 - val_acc: 0.7383
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0028 - acc: 0.7698Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0029 - acc: 0.7693 - val_loss: 0.0028 - val_acc: 0.7383
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0030 - acc: 0.7632Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7623 - val_loss: 0.0027 - val_acc: 0.7523
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0027 - acc: 0.7746Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7722 - val_loss: 0.0030 - val_acc: 0.7570
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0030 - acc: 0.7644Epoch 00031: val_loss improved from 0.00261 to 0.00246, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0030 - acc: 0.7629 - val_loss: 0.0025 - val_acc: 0.7523
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0025 - acc: 0.7632Epoch 00032: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7634 - val_loss: 0.0032 - val_acc: 0.7640
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0027 - acc: 0.7716Epoch 00033: val_loss improved from 0.00246 to 0.00176, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7745 - val_loss: 0.0018 - val_acc: 0.7710
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0028 - acc: 0.7764Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0028 - acc: 0.7769 - val_loss: 0.0024 - val_acc: 0.7664
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0025 - acc: 0.7686Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7681 - val_loss: 0.0037 - val_acc: 0.7664
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0027 - acc: 0.7626Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0027 - acc: 0.7652 - val_loss: 0.0029 - val_acc: 0.7687
Epoch 38/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0025 - acc: 0.7776Epoch 00037: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7769 - val_loss: 0.0020 - val_acc: 0.7664
Epoch 39/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0026 - acc: 0.7825Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0026 - acc: 0.7850 - val_loss: 0.0036 - val_acc: 0.7547
Epoch 40/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.7800Epoch 00039: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7780 - val_loss: 0.0052 - val_acc: 0.7617
Epoch 41/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0024 - acc: 0.7855Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7839 - val_loss: 0.0031 - val_acc: 0.7640
Epoch 42/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.7782Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0024 - acc: 0.7815 - val_loss: 0.0035 - val_acc: 0.7640
Epoch 43/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0022 - acc: 0.7861Epoch 00042: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7880 - val_loss: 0.0028 - val_acc: 0.7570
Epoch 44/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0022 - acc: 0.7867Epoch 00043: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7880 - val_loss: 0.0030 - val_acc: 0.7523
Epoch 45/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.7915Epoch 00044: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.7932 - val_loss: 0.0023 - val_acc: 0.7640
Epoch 46/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.7843Epoch 00045: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.7833 - val_loss: 0.0021 - val_acc: 0.7640
Epoch 47/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0022 - acc: 0.8143Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.8137 - val_loss: 0.0031 - val_acc: 0.7640
Epoch 48/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0021 - acc: 0.7927Epoch 00047: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7938 - val_loss: 0.0021 - val_acc: 0.7734
Epoch 49/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0026 - acc: 0.7957Epoch 00048: val_loss improved from 0.00176 to 0.00157, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7950 - val_loss: 0.0016 - val_acc: 0.7757
Epoch 50/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0020 - acc: 0.8053Epoch 00049: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.8043 - val_loss: 0.0018 - val_acc: 0.7734
Epoch 51/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0019 - acc: 0.8089Epoch 00050: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.8090 - val_loss: 0.0026 - val_acc: 0.7710
Epoch 52/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0021 - acc: 0.8005Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7985 - val_loss: 0.0017 - val_acc: 0.7710
Epoch 53/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0021 - acc: 0.7975Epoch 00052: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.7996 - val_loss: 0.0017 - val_acc: 0.7804
Epoch 54/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0020 - acc: 0.8053Epoch 00053: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.8061 - val_loss: 0.0029 - val_acc: 0.7734
Epoch 55/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8053Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.8072 - val_loss: 0.0023 - val_acc: 0.7804
Epoch 56/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.8101Epoch 00055: val_loss improved from 0.00157 to 0.00156, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0022 - acc: 0.8107 - val_loss: 0.0016 - val_acc: 0.7804
Epoch 57/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0020 - acc: 0.8113Epoch 00056: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.8084 - val_loss: 0.0017 - val_acc: 0.7710
Epoch 58/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8161Epoch 00057: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8154 - val_loss: 0.0020 - val_acc: 0.7757
Epoch 59/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8209Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.8224 - val_loss: 0.0025 - val_acc: 0.7710
Epoch 60/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8011Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7996 - val_loss: 0.0017 - val_acc: 0.7780
Epoch 61/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0020 - acc: 0.8101Epoch 00060: val_loss improved from 0.00156 to 0.00150, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.8113 - val_loss: 0.0015 - val_acc: 0.7710
Epoch 62/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0019 - acc: 0.8017Epoch 00061: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0019 - acc: 0.8002 - val_loss: 0.0019 - val_acc: 0.7734
Epoch 63/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8071Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8090 - val_loss: 0.0019 - val_acc: 0.7780
Epoch 64/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8155Epoch 00063: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.8172 - val_loss: 0.0018 - val_acc: 0.7780
Epoch 65/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8179Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8143 - val_loss: 0.0016 - val_acc: 0.7710
Epoch 66/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0023 - acc: 0.8053Epoch 00065: val_loss improved from 0.00150 to 0.00149, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0023 - acc: 0.8061 - val_loss: 0.0015 - val_acc: 0.7734
Epoch 67/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8221Epoch 00066: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8224 - val_loss: 0.0021 - val_acc: 0.7617
Epoch 68/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8155Epoch 00067: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.8154 - val_loss: 0.0017 - val_acc: 0.7850
Epoch 69/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0019 - acc: 0.8125Epoch 00068: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.8107 - val_loss: 0.0025 - val_acc: 0.7780
Epoch 70/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8233Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8189 - val_loss: 0.0015 - val_acc: 0.7874
Epoch 71/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8209Epoch 00070: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.8189 - val_loss: 0.0019 - val_acc: 0.7804
Epoch 72/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8215Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8189 - val_loss: 0.0021 - val_acc: 0.7827
Epoch 73/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8197Epoch 00072: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8189 - val_loss: 0.0016 - val_acc: 0.7687
Epoch 74/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0021 - acc: 0.8149Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0021 - acc: 0.8143 - val_loss: 0.0017 - val_acc: 0.7757
Epoch 75/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8287Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8271 - val_loss: 0.0019 - val_acc: 0.7804
Epoch 76/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8233Epoch 00075: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8254 - val_loss: 0.0021 - val_acc: 0.7734
Epoch 77/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8197Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.8218 - val_loss: 0.0025 - val_acc: 0.7710
Epoch 78/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8095Epoch 00077: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8107 - val_loss: 0.0019 - val_acc: 0.7734
Epoch 79/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8281Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8277 - val_loss: 0.0017 - val_acc: 0.7734
Epoch 80/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.8275Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.8289 - val_loss: 0.0026 - val_acc: 0.7780
Epoch 81/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8221Epoch 00080: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8248 - val_loss: 0.0019 - val_acc: 0.7757
Epoch 82/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8197Epoch 00081: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8189 - val_loss: 0.0022 - val_acc: 0.7734
Epoch 83/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.8233Epoch 00082: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8218 - val_loss: 0.0034 - val_acc: 0.7710
Epoch 84/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.8299Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.8283 - val_loss: 0.0036 - val_acc: 0.7780
Epoch 85/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8137Epoch 00084: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8137 - val_loss: 0.0018 - val_acc: 0.7850
Epoch 86/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8215Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8195 - val_loss: 0.0027 - val_acc: 0.7780
Epoch 87/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8263Epoch 00086: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8300 - val_loss: 0.0025 - val_acc: 0.7734
Epoch 88/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8329Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8306 - val_loss: 0.0016 - val_acc: 0.7664
Epoch 89/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8323Epoch 00088: val_loss improved from 0.00149 to 0.00143, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.8312 - val_loss: 0.0014 - val_acc: 0.7757
Epoch 90/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8257Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8248 - val_loss: 0.0028 - val_acc: 0.7640
Epoch 91/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8257Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8259 - val_loss: 0.0029 - val_acc: 0.7664
Epoch 92/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8275Epoch 00091: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8271 - val_loss: 0.0027 - val_acc: 0.7734
Epoch 93/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.8287Epoch 00092: val_loss improved from 0.00143 to 0.00137, saving model to my_model.rmsprop.h5
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8294 - val_loss: 0.0014 - val_acc: 0.7757
Epoch 94/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0017 - acc: 0.8245Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0017 - acc: 0.8265 - val_loss: 0.0017 - val_acc: 0.7850
Epoch 95/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8287Epoch 00094: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8283 - val_loss: 0.0019 - val_acc: 0.7804
Epoch 96/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.8474Epoch 00095: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.8470 - val_loss: 0.0018 - val_acc: 0.7780
Epoch 97/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.8329Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.8318 - val_loss: 0.0016 - val_acc: 0.7850
Epoch 98/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.8431Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.8446 - val_loss: 0.0014 - val_acc: 0.7850
Epoch 99/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0012 - acc: 0.8353Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.8324 - val_loss: 0.0023 - val_acc: 0.7687
Epoch 100/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.8323Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.8324 - val_loss: 0.0023 - val_acc: 0.7850
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.8377Epoch 00000: val_loss improved from inf to 0.00140, saving model to my_model.adam.h5
1712/1712 [==============================] - 1s - loss: 0.0011 - acc: 0.8411 - val_loss: 0.0014 - val_acc: 0.7850
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.1789e-04 - acc: 0.8444Epoch 00001: val_loss improved from 0.00140 to 0.00131, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 8.1927e-04 - acc: 0.8452 - val_loss: 0.0013 - val_acc: 0.7804
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7896e-04 - acc: 0.8329Epoch 00002: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7444e-04 - acc: 0.8359 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6771e-04 - acc: 0.8450Epoch 00003: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6576e-04 - acc: 0.8446 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4476e-04 - acc: 0.8377Epoch 00004: val_loss improved from 0.00131 to 0.00128, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 7.4686e-04 - acc: 0.8382 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4397e-04 - acc: 0.8335Epoch 00005: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4024e-04 - acc: 0.8359 - val_loss: 0.0013 - val_acc: 0.7804
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4054e-04 - acc: 0.8431Epoch 00006: val_loss improved from 0.00128 to 0.00125, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 7.3860e-04 - acc: 0.8440 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3229e-04 - acc: 0.8552Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3594e-04 - acc: 0.8551 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9428e-04 - acc: 0.8371Epoch 00008: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9641e-04 - acc: 0.8359 - val_loss: 0.0013 - val_acc: 0.7757
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0652e-04 - acc: 0.8522Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0414e-04 - acc: 0.8511 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0254e-04 - acc: 0.8444Epoch 00010: val_loss improved from 0.00125 to 0.00124, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 7.0407e-04 - acc: 0.8446 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.7535e-04 - acc: 0.8425Epoch 00011: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.7556e-04 - acc: 0.8429 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8992e-04 - acc: 0.8480Epoch 00012: val_loss improved from 0.00124 to 0.00121, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 6.8866e-04 - acc: 0.8464 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6895e-04 - acc: 0.8642Epoch 00013: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.6882e-04 - acc: 0.8633 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.4742e-04 - acc: 0.8672Epoch 00014: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.4521e-04 - acc: 0.8662 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.2657e-04 - acc: 0.8438Epoch 00015: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.2596e-04 - acc: 0.8464 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6168e-04 - acc: 0.8474Epoch 00016: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.5886e-04 - acc: 0.8487 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.5029e-04 - acc: 0.8504Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.5352e-04 - acc: 0.8481 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.4204e-04 - acc: 0.8425Epoch 00018: val_loss improved from 0.00121 to 0.00120, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 6.4241e-04 - acc: 0.8446 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.3714e-04 - acc: 0.8474Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.3578e-04 - acc: 0.8493 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.1642e-04 - acc: 0.8498Epoch 00020: val_loss improved from 0.00120 to 0.00119, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 6.1540e-04 - acc: 0.8475 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.1921e-04 - acc: 0.8570Epoch 00021: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.2114e-04 - acc: 0.8563 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.2695e-04 - acc: 0.8492Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.2648e-04 - acc: 0.8464 - val_loss: 0.0013 - val_acc: 0.7967
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.1627e-04 - acc: 0.8564Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.2285e-04 - acc: 0.8557 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.0168e-04 - acc: 0.8624Epoch 00024: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.0155e-04 - acc: 0.8633 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.2281e-04 - acc: 0.8570Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.2783e-04 - acc: 0.8581 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.1472e-04 - acc: 0.8419Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.1523e-04 - acc: 0.8440 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.0624e-04 - acc: 0.8582Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.0615e-04 - acc: 0.8581 - val_loss: 0.0012 - val_acc: 0.7757
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.0093e-04 - acc: 0.8582Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.0101e-04 - acc: 0.8598 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.9322e-04 - acc: 0.8654Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.9167e-04 - acc: 0.8657 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7443e-04 - acc: 0.8648Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.7311e-04 - acc: 0.8645 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.1487e-04 - acc: 0.8510Epoch 00031: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.1326e-04 - acc: 0.8534 - val_loss: 0.0014 - val_acc: 0.7874
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.4265e-04 - acc: 0.8486Epoch 00032: val_loss improved from 0.00119 to 0.00117, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 6.4010e-04 - acc: 0.8481 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.8987e-04 - acc: 0.8582Epoch 00033: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.8970e-04 - acc: 0.8586 - val_loss: 0.0012 - val_acc: 0.8061
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7826e-04 - acc: 0.8708Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.7770e-04 - acc: 0.8686 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7532e-04 - acc: 0.8630Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.7342e-04 - acc: 0.8633 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7527e-04 - acc: 0.8522Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.7906e-04 - acc: 0.8551 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 38/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.6202e-04 - acc: 0.8636Epoch 00037: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.6036e-04 - acc: 0.8639 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 39/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4770e-04 - acc: 0.8714Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4930e-04 - acc: 0.8738 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 40/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4722e-04 - acc: 0.8678Epoch 00039: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4921e-04 - acc: 0.8692 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 41/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7444e-04 - acc: 0.8528Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.7389e-04 - acc: 0.8522 - val_loss: 0.0012 - val_acc: 0.8014
Epoch 42/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4029e-04 - acc: 0.8492Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3908e-04 - acc: 0.8516 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 43/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3927e-04 - acc: 0.8642Epoch 00042: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3718e-04 - acc: 0.8639 - val_loss: 0.0012 - val_acc: 0.7804
Epoch 44/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3295e-04 - acc: 0.8588Epoch 00043: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3158e-04 - acc: 0.8569 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 45/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.2002e-04 - acc: 0.8612Epoch 00044: val_loss improved from 0.00117 to 0.00117, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 5.2139e-04 - acc: 0.8604 - val_loss: 0.0012 - val_acc: 0.8061
Epoch 46/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3153e-04 - acc: 0.8588Epoch 00045: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3107e-04 - acc: 0.8586 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 47/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4937e-04 - acc: 0.8660Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4978e-04 - acc: 0.8680 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 48/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3245e-04 - acc: 0.8624Epoch 00047: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3294e-04 - acc: 0.8604 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 49/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.1785e-04 - acc: 0.8558Epoch 00048: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.2195e-04 - acc: 0.8563 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 50/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.1947e-04 - acc: 0.8594Epoch 00049: val_loss improved from 0.00117 to 0.00115, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 5.1949e-04 - acc: 0.8616 - val_loss: 0.0012 - val_acc: 0.8061
Epoch 51/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3904e-04 - acc: 0.8444Epoch 00050: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4626e-04 - acc: 0.8458 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 52/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.2493e-04 - acc: 0.8540Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.2408e-04 - acc: 0.8540 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 53/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.2824e-04 - acc: 0.8738Epoch 00052: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.3096e-04 - acc: 0.8744 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 54/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.1991e-04 - acc: 0.8684Epoch 00053: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.2213e-04 - acc: 0.8692 - val_loss: 0.0012 - val_acc: 0.8014
Epoch 55/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4284e-04 - acc: 0.8750Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4171e-04 - acc: 0.8768 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 56/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.2907e-04 - acc: 0.8714Epoch 00055: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.2774e-04 - acc: 0.8721 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 57/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9535e-04 - acc: 0.8714Epoch 00056: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9554e-04 - acc: 0.8692 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 58/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0237e-04 - acc: 0.8606Epoch 00057: val_loss improved from 0.00115 to 0.00115, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 5.0283e-04 - acc: 0.8610 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 59/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0359e-04 - acc: 0.8678Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0042e-04 - acc: 0.8668 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 60/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0633e-04 - acc: 0.8654Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0840e-04 - acc: 0.8657 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 61/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.1488e-04 - acc: 0.8594Epoch 00060: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.1303e-04 - acc: 0.8598 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 62/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0196e-04 - acc: 0.8654Epoch 00061: val_loss improved from 0.00115 to 0.00113, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 5.0105e-04 - acc: 0.8668 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 63/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9488e-04 - acc: 0.8672Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9372e-04 - acc: 0.8680 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 64/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8230e-04 - acc: 0.8732Epoch 00063: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8393e-04 - acc: 0.8715 - val_loss: 0.0012 - val_acc: 0.8014
Epoch 65/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0979e-04 - acc: 0.8648Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0933e-04 - acc: 0.8651 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 66/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9806e-04 - acc: 0.8648Epoch 00065: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9871e-04 - acc: 0.8645 - val_loss: 0.0011 - val_acc: 0.7874
Epoch 67/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7873e-04 - acc: 0.8756Epoch 00066: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7704e-04 - acc: 0.8762 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 68/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6678e-04 - acc: 0.8672Epoch 00067: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6591e-04 - acc: 0.8662 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 69/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0081e-04 - acc: 0.8642Epoch 00068: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9917e-04 - acc: 0.8627 - val_loss: 0.0011 - val_acc: 0.7827
Epoch 70/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6631e-04 - acc: 0.8618Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6438e-04 - acc: 0.8592 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 71/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7235e-04 - acc: 0.8696Epoch 00070: val_loss improved from 0.00113 to 0.00113, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 4.7299e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.7780
Epoch 72/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7797e-04 - acc: 0.8804Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7748e-04 - acc: 0.8791 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 73/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7338e-04 - acc: 0.8696Epoch 00072: val_loss improved from 0.00113 to 0.00113, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 4.7247e-04 - acc: 0.8703 - val_loss: 0.0011 - val_acc: 0.7850
Epoch 74/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6365e-04 - acc: 0.8732Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6396e-04 - acc: 0.8744 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 75/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7686e-04 - acc: 0.8714Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7639e-04 - acc: 0.8721 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 76/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6430e-04 - acc: 0.8744Epoch 00075: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6391e-04 - acc: 0.8738 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 77/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8188e-04 - acc: 0.8672Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8323e-04 - acc: 0.8680 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 78/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0053e-04 - acc: 0.8552Epoch 00077: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0184e-04 - acc: 0.8569 - val_loss: 0.0011 - val_acc: 0.7874
Epoch 79/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8225e-04 - acc: 0.8624Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8236e-04 - acc: 0.8639 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 80/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6440e-04 - acc: 0.8594Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6342e-04 - acc: 0.8598 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 81/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6947e-04 - acc: 0.8642Epoch 00080: val_loss improved from 0.00113 to 0.00111, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 4.6959e-04 - acc: 0.8645 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 82/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6394e-04 - acc: 0.8648Epoch 00081: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6388e-04 - acc: 0.8662 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 83/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7360e-04 - acc: 0.8720Epoch 00082: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7415e-04 - acc: 0.8732 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 84/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7099e-04 - acc: 0.8732Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7243e-04 - acc: 0.8727 - val_loss: 0.0011 - val_acc: 0.7921
Epoch 85/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6593e-04 - acc: 0.8774Epoch 00084: val_loss improved from 0.00111 to 0.00111, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 4.6633e-04 - acc: 0.8773 - val_loss: 0.0011 - val_acc: 0.7827
Epoch 86/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6348e-04 - acc: 0.8744Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6297e-04 - acc: 0.8744 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 87/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5840e-04 - acc: 0.8576Epoch 00086: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5942e-04 - acc: 0.8581 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 88/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6847e-04 - acc: 0.8672Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6916e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 89/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5514e-04 - acc: 0.8702Epoch 00088: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5891e-04 - acc: 0.8692 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 90/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6289e-04 - acc: 0.8684Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6205e-04 - acc: 0.8697 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 91/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5147e-04 - acc: 0.8678Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5166e-04 - acc: 0.8697 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 92/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4686e-04 - acc: 0.8672Epoch 00091: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4531e-04 - acc: 0.8662 - val_loss: 0.0011 - val_acc: 0.7827
Epoch 93/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7807e-04 - acc: 0.8720Epoch 00092: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7713e-04 - acc: 0.8703 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 94/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7825e-04 - acc: 0.8678Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7992e-04 - acc: 0.8674 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 95/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4825e-04 - acc: 0.8750Epoch 00094: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4822e-04 - acc: 0.8727 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 96/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4394e-04 - acc: 0.8750Epoch 00095: val_loss improved from 0.00111 to 0.00110, saving model to my_model.adam.h5
1712/1712 [==============================] - 0s - loss: 4.4595e-04 - acc: 0.8750 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 97/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.3851e-04 - acc: 0.8612Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.3716e-04 - acc: 0.8610 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 98/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.3600e-04 - acc: 0.8690Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.3698e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 99/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6284e-04 - acc: 0.8708Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6389e-04 - acc: 0.8721 - val_loss: 0.0011 - val_acc: 0.7804
Epoch 100/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4995e-04 - acc: 0.8798Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4950e-04 - acc: 0.8797 - val_loss: 0.0011 - val_acc: 0.7874
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1483e-04 - acc: 0.8804Epoch 00000: val_loss improved from inf to 0.00111, saving model to my_model.sgd.h5
1712/1712 [==============================] - 0s - loss: 4.1428e-04 - acc: 0.8808 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2146e-04 - acc: 0.8744Epoch 00001: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2161e-04 - acc: 0.8738 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1960e-04 - acc: 0.8798Epoch 00002: val_loss improved from 0.00111 to 0.00111, saving model to my_model.sgd.h5
1712/1712 [==============================] - 0s - loss: 4.2060e-04 - acc: 0.8797 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1950e-04 - acc: 0.8792Epoch 00003: val_loss improved from 0.00111 to 0.00111, saving model to my_model.sgd.h5
1712/1712 [==============================] - 0s - loss: 4.1925e-04 - acc: 0.8803 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2173e-04 - acc: 0.8744Epoch 00004: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2010e-04 - acc: 0.8744 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1774e-04 - acc: 0.8648Epoch 00005: val_loss improved from 0.00111 to 0.00110, saving model to my_model.sgd.h5
1712/1712 [==============================] - 0s - loss: 4.1673e-04 - acc: 0.8645 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0843e-04 - acc: 0.8804Epoch 00006: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1025e-04 - acc: 0.8779 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2342e-04 - acc: 0.8750Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2390e-04 - acc: 0.8744 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1767e-04 - acc: 0.8786Epoch 00008: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1897e-04 - acc: 0.8785 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1863e-04 - acc: 0.8828Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1697e-04 - acc: 0.8814 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1914e-04 - acc: 0.8792Epoch 00010: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1910e-04 - acc: 0.8785 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2563e-04 - acc: 0.8858Epoch 00011: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2458e-04 - acc: 0.8849 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1805e-04 - acc: 0.8708Epoch 00012: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1817e-04 - acc: 0.8686 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1354e-04 - acc: 0.8780Epoch 00013: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1466e-04 - acc: 0.8768 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2251e-04 - acc: 0.8636Epoch 00014: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2193e-04 - acc: 0.8633 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1578e-04 - acc: 0.8720Epoch 00015: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1580e-04 - acc: 0.8732 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1562e-04 - acc: 0.8708Epoch 00016: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1691e-04 - acc: 0.8709 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1697e-04 - acc: 0.8840Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1660e-04 - acc: 0.8832 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1973e-04 - acc: 0.8738Epoch 00018: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1953e-04 - acc: 0.8762 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0820e-04 - acc: 0.8798Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0789e-04 - acc: 0.8797 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1115e-04 - acc: 0.8702Epoch 00020: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1103e-04 - acc: 0.8709 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.3039e-04 - acc: 0.8714Epoch 00021: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2962e-04 - acc: 0.8727 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1948e-04 - acc: 0.8738Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2003e-04 - acc: 0.8738 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1403e-04 - acc: 0.8846Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1389e-04 - acc: 0.8826 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1605e-04 - acc: 0.8660Epoch 00024: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1621e-04 - acc: 0.8674 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1096e-04 - acc: 0.8774Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1144e-04 - acc: 0.8768 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1426e-04 - acc: 0.8750Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1418e-04 - acc: 0.8768 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2116e-04 - acc: 0.8786Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2272e-04 - acc: 0.8814 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0659e-04 - acc: 0.8786Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0947e-04 - acc: 0.8768 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1412e-04 - acc: 0.8750Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1450e-04 - acc: 0.8738 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1839e-04 - acc: 0.8648Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1875e-04 - acc: 0.8627 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2257e-04 - acc: 0.8756Epoch 00031: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2452e-04 - acc: 0.8732 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2061e-04 - acc: 0.8678Epoch 00032: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1931e-04 - acc: 0.8697 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1327e-04 - acc: 0.8714Epoch 00033: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1251e-04 - acc: 0.8727 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1600e-04 - acc: 0.8774Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1570e-04 - acc: 0.8785 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0819e-04 - acc: 0.8888Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0875e-04 - acc: 0.8879 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0944e-04 - acc: 0.8642Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1016e-04 - acc: 0.8645 - val_loss: 0.0011 - val_acc: 0.7921
Epoch 00036: early stopping
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0589 - acc: 0.6977Epoch 00000: val_loss improved from inf to 0.00383, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0573 - acc: 0.6980 - val_loss: 0.0038 - val_acc: 0.6963
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0036 - acc: 0.7188Epoch 00001: val_loss improved from 0.00383 to 0.00288, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0036 - acc: 0.7155 - val_loss: 0.0029 - val_acc: 0.7173
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0025 - acc: 0.7344Epoch 00002: val_loss improved from 0.00288 to 0.00215, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0025 - acc: 0.7377 - val_loss: 0.0022 - val_acc: 0.7523
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0020 - acc: 0.7464Epoch 00003: val_loss improved from 0.00215 to 0.00210, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0020 - acc: 0.7471 - val_loss: 0.0021 - val_acc: 0.7360
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0018 - acc: 0.7596Epoch 00004: val_loss improved from 0.00210 to 0.00172, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0018 - acc: 0.7564 - val_loss: 0.0017 - val_acc: 0.7547
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0016 - acc: 0.7698Epoch 00005: val_loss improved from 0.00172 to 0.00164, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0016 - acc: 0.7699 - val_loss: 0.0016 - val_acc: 0.7570
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0015 - acc: 0.7650Epoch 00006: val_loss improved from 0.00164 to 0.00163, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0015 - acc: 0.7623 - val_loss: 0.0016 - val_acc: 0.7617
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.7728Epoch 00007: val_loss improved from 0.00163 to 0.00161, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7716 - val_loss: 0.0016 - val_acc: 0.7757
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.7722Epoch 00008: val_loss improved from 0.00161 to 0.00157, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7710 - val_loss: 0.0016 - val_acc: 0.7593
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0014 - acc: 0.7770Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0014 - acc: 0.7757 - val_loss: 0.0016 - val_acc: 0.7640
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.7752Epoch 00010: val_loss improved from 0.00157 to 0.00156, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7757 - val_loss: 0.0016 - val_acc: 0.7593
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0013 - acc: 0.7770Epoch 00011: val_loss improved from 0.00156 to 0.00149, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0013 - acc: 0.7775 - val_loss: 0.0015 - val_acc: 0.7570
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0012 - acc: 0.7849Epoch 00012: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7845 - val_loss: 0.0015 - val_acc: 0.7617
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0012 - acc: 0.7879Epoch 00013: val_loss improved from 0.00149 to 0.00147, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7880 - val_loss: 0.0015 - val_acc: 0.7593
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0012 - acc: 0.7831Epoch 00014: val_loss improved from 0.00147 to 0.00146, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0012 - acc: 0.7856 - val_loss: 0.0015 - val_acc: 0.7664
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7909Epoch 00015: val_loss improved from 0.00146 to 0.00143, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7909 - val_loss: 0.0014 - val_acc: 0.7687
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7819Epoch 00016: val_loss improved from 0.00143 to 0.00141, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7821 - val_loss: 0.0014 - val_acc: 0.7734
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7800Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7821 - val_loss: 0.0014 - val_acc: 0.7640
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7987Epoch 00018: val_loss improved from 0.00141 to 0.00139, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7973 - val_loss: 0.0014 - val_acc: 0.7757
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7921Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7926 - val_loss: 0.0014 - val_acc: 0.7734
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7885Epoch 00020: val_loss improved from 0.00139 to 0.00137, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7903 - val_loss: 0.0014 - val_acc: 0.7664
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7951Epoch 00021: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0011 - acc: 0.7956 - val_loss: 0.0014 - val_acc: 0.7757
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0011 - acc: 0.7999Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0010 - acc: 0.8014 - val_loss: 0.0014 - val_acc: 0.7687
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0010 - acc: 0.7740Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 0.0010 - acc: 0.7757 - val_loss: 0.0014 - val_acc: 0.7734
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0010 - acc: 0.7873Epoch 00024: val_loss improved from 0.00137 to 0.00136, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 0.0010 - acc: 0.7868 - val_loss: 0.0014 - val_acc: 0.7710
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0010 - acc: 0.8053 Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.9680e-04 - acc: 0.8067 - val_loss: 0.0014 - val_acc: 0.7664
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.9530e-04 - acc: 0.7927Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.9597e-04 - acc: 0.7932 - val_loss: 0.0014 - val_acc: 0.7664
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.9446e-04 - acc: 0.8149Epoch 00027: val_loss improved from 0.00136 to 0.00135, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.9273e-04 - acc: 0.8143 - val_loss: 0.0013 - val_acc: 0.7687
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.9034e-04 - acc: 0.8047Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.8850e-04 - acc: 0.8061 - val_loss: 0.0014 - val_acc: 0.7664
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.7977e-04 - acc: 0.7987Epoch 00029: val_loss improved from 0.00135 to 0.00134, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.7563e-04 - acc: 0.8008 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.5644e-04 - acc: 0.8167Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.5865e-04 - acc: 0.8160 - val_loss: 0.0014 - val_acc: 0.7734
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.6134e-04 - acc: 0.7981Epoch 00031: val_loss improved from 0.00134 to 0.00133, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.5924e-04 - acc: 0.7991 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.4314e-04 - acc: 0.8059Epoch 00032: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.4134e-04 - acc: 0.8067 - val_loss: 0.0014 - val_acc: 0.7640
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.3695e-04 - acc: 0.8023Epoch 00033: val_loss improved from 0.00133 to 0.00131, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.3789e-04 - acc: 0.8026 - val_loss: 0.0013 - val_acc: 0.7593
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.2833e-04 - acc: 0.8023Epoch 00034: val_loss improved from 0.00131 to 0.00131, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.2509e-04 - acc: 0.8032 - val_loss: 0.0013 - val_acc: 0.7593
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.1047e-04 - acc: 0.8023Epoch 00035: val_loss improved from 0.00131 to 0.00130, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 9.1059e-04 - acc: 0.8020 - val_loss: 0.0013 - val_acc: 0.7617
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.2229e-04 - acc: 0.8101Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.2634e-04 - acc: 0.8096 - val_loss: 0.0014 - val_acc: 0.7640
Epoch 38/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.2069e-04 - acc: 0.8071Epoch 00037: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.2293e-04 - acc: 0.8067 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 39/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.8180e-04 - acc: 0.8089Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.8111e-04 - acc: 0.8090 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 40/100
1664/1712 [============================>.] - ETA: 0s - loss: 9.0785e-04 - acc: 0.8029Epoch 00039: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 9.0656e-04 - acc: 0.8032 - val_loss: 0.0013 - val_acc: 0.7710
Epoch 41/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.9314e-04 - acc: 0.8023Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.9022e-04 - acc: 0.8032 - val_loss: 0.0013 - val_acc: 0.7687
Epoch 42/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.7509e-04 - acc: 0.8131Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.7382e-04 - acc: 0.8148 - val_loss: 0.0013 - val_acc: 0.7687
Epoch 43/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.8275e-04 - acc: 0.8191Epoch 00042: val_loss improved from 0.00130 to 0.00127, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 8.8368e-04 - acc: 0.8166 - val_loss: 0.0013 - val_acc: 0.7780
Epoch 44/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.8012e-04 - acc: 0.8137Epoch 00043: val_loss improved from 0.00127 to 0.00127, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 8.8379e-04 - acc: 0.8137 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 45/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.6808e-04 - acc: 0.8215Epoch 00044: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.7457e-04 - acc: 0.8201 - val_loss: 0.0013 - val_acc: 0.7780
Epoch 46/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.6589e-04 - acc: 0.8143Epoch 00045: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.6490e-04 - acc: 0.8125 - val_loss: 0.0013 - val_acc: 0.7780
Epoch 47/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.7257e-04 - acc: 0.8149Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.7127e-04 - acc: 0.8137 - val_loss: 0.0013 - val_acc: 0.7780
Epoch 48/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.4741e-04 - acc: 0.8167Epoch 00047: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.4222e-04 - acc: 0.8166 - val_loss: 0.0013 - val_acc: 0.7734
Epoch 49/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.7841e-04 - acc: 0.8011Epoch 00048: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.7704e-04 - acc: 0.8014 - val_loss: 0.0014 - val_acc: 0.7827
Epoch 50/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.6068e-04 - acc: 0.8179Epoch 00049: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.5909e-04 - acc: 0.8189 - val_loss: 0.0013 - val_acc: 0.7804
Epoch 51/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.4395e-04 - acc: 0.8149Epoch 00050: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.4036e-04 - acc: 0.8160 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 52/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.5156e-04 - acc: 0.8071Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.5353e-04 - acc: 0.8072 - val_loss: 0.0013 - val_acc: 0.7710
Epoch 53/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.5913e-04 - acc: 0.8167Epoch 00052: val_loss improved from 0.00127 to 0.00126, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 8.5773e-04 - acc: 0.8172 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 54/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.3165e-04 - acc: 0.8107Epoch 00053: val_loss improved from 0.00126 to 0.00125, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 8.2748e-04 - acc: 0.8119 - val_loss: 0.0013 - val_acc: 0.7734
Epoch 55/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.4489e-04 - acc: 0.8071Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.4150e-04 - acc: 0.8096 - val_loss: 0.0013 - val_acc: 0.7757
Epoch 56/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.3524e-04 - acc: 0.8089Epoch 00055: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.3344e-04 - acc: 0.8107 - val_loss: 0.0013 - val_acc: 0.7664
Epoch 57/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.3340e-04 - acc: 0.8023Epoch 00056: val_loss improved from 0.00125 to 0.00124, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 8.4427e-04 - acc: 0.8002 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 58/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.0739e-04 - acc: 0.8125Epoch 00057: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.0847e-04 - acc: 0.8143 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 59/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.2354e-04 - acc: 0.8017Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.1918e-04 - acc: 0.8037 - val_loss: 0.0013 - val_acc: 0.7757
Epoch 60/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.2644e-04 - acc: 0.8053Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.2377e-04 - acc: 0.8049 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 61/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.2937e-04 - acc: 0.8143Epoch 00060: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.3064e-04 - acc: 0.8148 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 62/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.1621e-04 - acc: 0.8233Epoch 00061: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.1393e-04 - acc: 0.8248 - val_loss: 0.0013 - val_acc: 0.7757
Epoch 63/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.1466e-04 - acc: 0.8179Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.0973e-04 - acc: 0.8172 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 64/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8456e-04 - acc: 0.8167Epoch 00063: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9238e-04 - acc: 0.8166 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 65/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.1256e-04 - acc: 0.8245Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.0727e-04 - acc: 0.8271 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 66/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8560e-04 - acc: 0.8191Epoch 00065: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.8921e-04 - acc: 0.8189 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 67/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8436e-04 - acc: 0.8221Epoch 00066: val_loss improved from 0.00124 to 0.00123, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.8572e-04 - acc: 0.8230 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 68/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.0217e-04 - acc: 0.8053Epoch 00067: val_loss improved from 0.00123 to 0.00123, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.9916e-04 - acc: 0.8072 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 69/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8963e-04 - acc: 0.8251Epoch 00068: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.8783e-04 - acc: 0.8271 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 70/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9714e-04 - acc: 0.8179Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9789e-04 - acc: 0.8166 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 71/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9337e-04 - acc: 0.8095Epoch 00070: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9392e-04 - acc: 0.8107 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 72/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9794e-04 - acc: 0.8149Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9658e-04 - acc: 0.8148 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 73/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5817e-04 - acc: 0.8047Epoch 00072: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5781e-04 - acc: 0.8049 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 74/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7607e-04 - acc: 0.8215Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7076e-04 - acc: 0.8201 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 75/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9613e-04 - acc: 0.8179Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9630e-04 - acc: 0.8160 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 76/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8046e-04 - acc: 0.8311Epoch 00075: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.8109e-04 - acc: 0.8324 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 77/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5116e-04 - acc: 0.8323Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5319e-04 - acc: 0.8329 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 78/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6784e-04 - acc: 0.8197Epoch 00077: val_loss improved from 0.00123 to 0.00122, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.6780e-04 - acc: 0.8183 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 79/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6630e-04 - acc: 0.8275Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6558e-04 - acc: 0.8259 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 80/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5997e-04 - acc: 0.8347Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6176e-04 - acc: 0.8376 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 81/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7625e-04 - acc: 0.8221Epoch 00080: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7322e-04 - acc: 0.8213 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 82/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6721e-04 - acc: 0.8155Epoch 00081: val_loss improved from 0.00122 to 0.00122, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.6240e-04 - acc: 0.8172 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 83/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4913e-04 - acc: 0.8143Epoch 00082: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5438e-04 - acc: 0.8148 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 84/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3583e-04 - acc: 0.8329Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4093e-04 - acc: 0.8324 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 85/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3594e-04 - acc: 0.8377Epoch 00084: val_loss improved from 0.00122 to 0.00121, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.3450e-04 - acc: 0.8347 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 86/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5238e-04 - acc: 0.8227Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4870e-04 - acc: 0.8230 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 87/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5391e-04 - acc: 0.8287Epoch 00086: val_loss improved from 0.00121 to 0.00121, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.5602e-04 - acc: 0.8289 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 88/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4852e-04 - acc: 0.8311Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5025e-04 - acc: 0.8294 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 89/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4037e-04 - acc: 0.8221Epoch 00088: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3824e-04 - acc: 0.8218 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 90/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4366e-04 - acc: 0.8197Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3864e-04 - acc: 0.8218 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 91/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3452e-04 - acc: 0.8239Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3230e-04 - acc: 0.8254 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 92/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4435e-04 - acc: 0.8167Epoch 00091: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4326e-04 - acc: 0.8178 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 93/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2326e-04 - acc: 0.8185Epoch 00092: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2113e-04 - acc: 0.8207 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 94/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1539e-04 - acc: 0.8323Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1655e-04 - acc: 0.8353 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 95/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3040e-04 - acc: 0.8155Epoch 00094: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3411e-04 - acc: 0.8178 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 96/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2979e-04 - acc: 0.8293Epoch 00095: val_loss improved from 0.00121 to 0.00120, saving model to my_model.adagrad.h5
1712/1712 [==============================] - 0s - loss: 7.3197e-04 - acc: 0.8283 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 97/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2640e-04 - acc: 0.8287Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2512e-04 - acc: 0.8283 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 98/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2888e-04 - acc: 0.8203Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2959e-04 - acc: 0.8224 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 99/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0755e-04 - acc: 0.8263Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0779e-04 - acc: 0.8277 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 100/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2081e-04 - acc: 0.8257Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2050e-04 - acc: 0.8277 - val_loss: 0.0012 - val_acc: 0.7897
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.2877e-04 - acc: 0.8299Epoch 00000: val_loss improved from inf to 0.00157, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 1s - loss: 8.2917e-04 - acc: 0.8324 - val_loss: 0.0016 - val_acc: 0.7850
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5583e-04 - acc: 0.8251Epoch 00001: val_loss improved from 0.00157 to 0.00128, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.5404e-04 - acc: 0.8259 - val_loss: 0.0013 - val_acc: 0.7991
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3154e-04 - acc: 0.8311Epoch 00002: val_loss improved from 0.00128 to 0.00121, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.3276e-04 - acc: 0.8306 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5133e-04 - acc: 0.8125Epoch 00003: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5864e-04 - acc: 0.8107 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3353e-04 - acc: 0.8191Epoch 00004: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3136e-04 - acc: 0.8207 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8406e-04 - acc: 0.8317Epoch 00005: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.8452e-04 - acc: 0.8324 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.8187e-04 - acc: 0.8305Epoch 00006: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.8173e-04 - acc: 0.8306 - val_loss: 0.0014 - val_acc: 0.7897
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9788e-04 - acc: 0.8269Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.0234e-04 - acc: 0.8283 - val_loss: 0.0014 - val_acc: 0.7967
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5111e-04 - acc: 0.8197Epoch 00008: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5015e-04 - acc: 0.8207 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4170e-04 - acc: 0.8167Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5169e-04 - acc: 0.8172 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 8.5914e-04 - acc: 0.8143Epoch 00010: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 8.5633e-04 - acc: 0.8131 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4812e-04 - acc: 0.8233Epoch 00011: val_loss improved from 0.00121 to 0.00120, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.4790e-04 - acc: 0.8242 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2065e-04 - acc: 0.8245Epoch 00012: val_loss improved from 0.00120 to 0.00120, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.1912e-04 - acc: 0.8259 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2118e-04 - acc: 0.8353Epoch 00013: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2379e-04 - acc: 0.8353 - val_loss: 0.0013 - val_acc: 0.7991
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1513e-04 - acc: 0.8245Epoch 00014: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1044e-04 - acc: 0.8265 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1992e-04 - acc: 0.8395Epoch 00015: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1959e-04 - acc: 0.8400 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6096e-04 - acc: 0.8275Epoch 00016: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6228e-04 - acc: 0.8306 - val_loss: 0.0017 - val_acc: 0.7850
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9759e-04 - acc: 0.8209Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9369e-04 - acc: 0.8218 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1674e-04 - acc: 0.8419Epoch 00018: val_loss improved from 0.00120 to 0.00119, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.1911e-04 - acc: 0.8405 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3024e-04 - acc: 0.8311Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3180e-04 - acc: 0.8318 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3583e-04 - acc: 0.8191Epoch 00020: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3887e-04 - acc: 0.8195 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7800e-04 - acc: 0.8239Epoch 00021: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7877e-04 - acc: 0.8207 - val_loss: 0.0013 - val_acc: 0.7991
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0992e-04 - acc: 0.8257Epoch 00022: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1157e-04 - acc: 0.8259 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1339e-04 - acc: 0.8287Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1138e-04 - acc: 0.8300 - val_loss: 0.0012 - val_acc: 0.7991
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4434e-04 - acc: 0.8239Epoch 00024: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4330e-04 - acc: 0.8254 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0249e-04 - acc: 0.8209Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0232e-04 - acc: 0.8213 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1049e-04 - acc: 0.8263Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1170e-04 - acc: 0.8254 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3819e-04 - acc: 0.8239Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4140e-04 - acc: 0.8218 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7280e-04 - acc: 0.8281Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7045e-04 - acc: 0.8259 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1285e-04 - acc: 0.8335Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1359e-04 - acc: 0.8335 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1466e-04 - acc: 0.8353Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1487e-04 - acc: 0.8353 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2825e-04 - acc: 0.8233Epoch 00031: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2740e-04 - acc: 0.8224 - val_loss: 0.0014 - val_acc: 0.7897
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9275e-04 - acc: 0.8359Epoch 00032: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9034e-04 - acc: 0.8364 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2518e-04 - acc: 0.8287Epoch 00033: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2345e-04 - acc: 0.8318 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6691e-04 - acc: 0.8233Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6214e-04 - acc: 0.8254 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2314e-04 - acc: 0.8438Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1920e-04 - acc: 0.8452 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9288e-04 - acc: 0.8251Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9274e-04 - acc: 0.8277 - val_loss: 0.0014 - val_acc: 0.7850
Epoch 38/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.6949e-04 - acc: 0.8353Epoch 00037: val_loss improved from 0.00119 to 0.00119, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.6457e-04 - acc: 0.8353 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 39/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2255e-04 - acc: 0.8239Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2176e-04 - acc: 0.8254 - val_loss: 0.0013 - val_acc: 0.7827
Epoch 40/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1980e-04 - acc: 0.8293Epoch 00039: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2582e-04 - acc: 0.8283 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 41/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5946e-04 - acc: 0.8275Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.6214e-04 - acc: 0.8265 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 42/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0322e-04 - acc: 0.8383Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0164e-04 - acc: 0.8388 - val_loss: 0.0014 - val_acc: 0.7921
Epoch 43/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2728e-04 - acc: 0.8305Epoch 00042: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2334e-04 - acc: 0.8294 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 44/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9225e-04 - acc: 0.8341Epoch 00043: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9392e-04 - acc: 0.8329 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 45/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4781e-04 - acc: 0.8377Epoch 00044: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4587e-04 - acc: 0.8364 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 46/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4152e-04 - acc: 0.8371Epoch 00045: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3931e-04 - acc: 0.8364 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 47/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0949e-04 - acc: 0.8257Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0678e-04 - acc: 0.8271 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 48/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8951e-04 - acc: 0.8233Epoch 00047: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9143e-04 - acc: 0.8242 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 49/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5621e-04 - acc: 0.8329Epoch 00048: val_loss improved from 0.00119 to 0.00119, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.5155e-04 - acc: 0.8347 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 50/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0973e-04 - acc: 0.8341Epoch 00049: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1055e-04 - acc: 0.8347 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 51/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1150e-04 - acc: 0.8293Epoch 00050: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1003e-04 - acc: 0.8306 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 52/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2526e-04 - acc: 0.8365Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2927e-04 - acc: 0.8364 - val_loss: 0.0014 - val_acc: 0.7921
Epoch 53/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3142e-04 - acc: 0.8311Epoch 00052: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3059e-04 - acc: 0.8300 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 54/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.7245e-04 - acc: 0.8438Epoch 00053: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.7078e-04 - acc: 0.8446 - val_loss: 0.0012 - val_acc: 0.7967
Epoch 55/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1036e-04 - acc: 0.8383Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1010e-04 - acc: 0.8388 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 56/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.9776e-04 - acc: 0.8335Epoch 00055: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.9680e-04 - acc: 0.8329 - val_loss: 0.0014 - val_acc: 0.7921
Epoch 57/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5252e-04 - acc: 0.8413Epoch 00056: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4936e-04 - acc: 0.8429 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 58/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8347e-04 - acc: 0.8359Epoch 00057: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8002e-04 - acc: 0.8370 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 59/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9827e-04 - acc: 0.8245Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9803e-04 - acc: 0.8254 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 60/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3203e-04 - acc: 0.8480Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.3325e-04 - acc: 0.8429 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 61/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8647e-04 - acc: 0.8359Epoch 00060: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8536e-04 - acc: 0.8359 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 62/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9336e-04 - acc: 0.8353Epoch 00061: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9247e-04 - acc: 0.8359 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 63/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0368e-04 - acc: 0.8305Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0248e-04 - acc: 0.8312 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 64/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9093e-04 - acc: 0.8401Epoch 00063: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9549e-04 - acc: 0.8388 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 65/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5706e-04 - acc: 0.8293Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5625e-04 - acc: 0.8289 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 66/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4420e-04 - acc: 0.8311Epoch 00065: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.4154e-04 - acc: 0.8312 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 67/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0211e-04 - acc: 0.8359Epoch 00066: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0148e-04 - acc: 0.8370 - val_loss: 0.0013 - val_acc: 0.7944
Epoch 68/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0975e-04 - acc: 0.8299Epoch 00067: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1082e-04 - acc: 0.8324 - val_loss: 0.0013 - val_acc: 0.7967
Epoch 69/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9628e-04 - acc: 0.8233Epoch 00068: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9390e-04 - acc: 0.8248 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 70/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8125e-04 - acc: 0.8257Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8435e-04 - acc: 0.8254 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 71/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4536e-04 - acc: 0.8329Epoch 00070: val_loss improved from 0.00119 to 0.00118, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 7.4531e-04 - acc: 0.8341 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 72/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.7749e-04 - acc: 0.8263Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8379e-04 - acc: 0.8254 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 73/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8158e-04 - acc: 0.8438Epoch 00072: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8277e-04 - acc: 0.8435 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 74/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.1692e-04 - acc: 0.8359Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.1613e-04 - acc: 0.8364 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 75/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9505e-04 - acc: 0.8383Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9202e-04 - acc: 0.8405 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 76/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.7256e-04 - acc: 0.8365Epoch 00075: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.7557e-04 - acc: 0.8370 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 77/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9068e-04 - acc: 0.8269Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8905e-04 - acc: 0.8283 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 78/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9169e-04 - acc: 0.8389Epoch 00077: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9112e-04 - acc: 0.8400 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 79/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6304e-04 - acc: 0.8419Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.6294e-04 - acc: 0.8411 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 80/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2313e-04 - acc: 0.8401Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2509e-04 - acc: 0.8405 - val_loss: 0.0014 - val_acc: 0.7897
Epoch 81/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.2608e-04 - acc: 0.8359Epoch 00080: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2320e-04 - acc: 0.8382 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 82/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.4697e-04 - acc: 0.8365Epoch 00081: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5063e-04 - acc: 0.8335 - val_loss: 0.0012 - val_acc: 0.7944
Epoch 83/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8233e-04 - acc: 0.8317Epoch 00082: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8215e-04 - acc: 0.8306 - val_loss: 0.0013 - val_acc: 0.7897
Epoch 84/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8370e-04 - acc: 0.8311Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8410e-04 - acc: 0.8324 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 85/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.5836e-04 - acc: 0.8329Epoch 00084: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.5814e-04 - acc: 0.8329 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 86/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9786e-04 - acc: 0.8221Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0112e-04 - acc: 0.8242 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 87/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8813e-04 - acc: 0.8371Epoch 00086: val_loss improved from 0.00118 to 0.00117, saving model to my_model.adadelta.h5
1712/1712 [==============================] - 0s - loss: 6.8786e-04 - acc: 0.8359 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 88/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8004e-04 - acc: 0.8077Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.9159e-04 - acc: 0.8096 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 89/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.0440e-04 - acc: 0.8269Epoch 00088: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0192e-04 - acc: 0.8271 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 90/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8390e-04 - acc: 0.8299Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8600e-04 - acc: 0.8300 - val_loss: 0.0013 - val_acc: 0.7874
Epoch 91/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.7330e-04 - acc: 0.8462Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.7365e-04 - acc: 0.8435 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 92/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.5437e-04 - acc: 0.8263Epoch 00091: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.5826e-04 - acc: 0.8265 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 93/100
1664/1712 [============================>.] - ETA: 0s - loss: 7.3228e-04 - acc: 0.8450Epoch 00092: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.2853e-04 - acc: 0.8458 - val_loss: 0.0013 - val_acc: 0.7850
Epoch 94/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.9864e-04 - acc: 0.8419Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 7.0373e-04 - acc: 0.8435 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 95/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8469e-04 - acc: 0.8299Epoch 00094: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8678e-04 - acc: 0.8289 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 96/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8903e-04 - acc: 0.8371Epoch 00095: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.8705e-04 - acc: 0.8394 - val_loss: 0.0012 - val_acc: 0.7897
Epoch 97/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.7219e-04 - acc: 0.8203Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.7063e-04 - acc: 0.8201 - val_loss: 0.0012 - val_acc: 0.7921
Epoch 98/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6171e-04 - acc: 0.8299Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.6122e-04 - acc: 0.8324 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 99/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6202e-04 - acc: 0.8407Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.6384e-04 - acc: 0.8388 - val_loss: 0.0012 - val_acc: 0.7874
Epoch 100/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.6392e-04 - acc: 0.8263Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 6.6036e-04 - acc: 0.8294 - val_loss: 0.0013 - val_acc: 0.7921
Train on 1712 samples, validate on 428 samples
Epoch 1/100
1664/1712 [============================>.] - ETA: 0s - loss: 0.0010 - acc: 0.8155Epoch 00000: val_loss improved from inf to 0.00126, saving model to my_model.adamax.h5
1712/1712 [==============================] - 1s - loss: 0.0010 - acc: 0.8160 - val_loss: 0.0013 - val_acc: 0.7921
Epoch 2/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.8569e-04 - acc: 0.8353Epoch 00001: val_loss improved from 0.00126 to 0.00122, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 6.8572e-04 - acc: 0.8370 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 3/100
1664/1712 [============================>.] - ETA: 0s - loss: 6.3642e-04 - acc: 0.8365Epoch 00002: val_loss improved from 0.00122 to 0.00119, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 6.3458e-04 - acc: 0.8359 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 4/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.9961e-04 - acc: 0.8395Epoch 00003: val_loss improved from 0.00119 to 0.00117, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.9833e-04 - acc: 0.8394 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 5/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.8393e-04 - acc: 0.8323Epoch 00004: val_loss improved from 0.00117 to 0.00116, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.8193e-04 - acc: 0.8335 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 6/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.8790e-04 - acc: 0.8227Epoch 00005: val_loss improved from 0.00116 to 0.00115, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.8650e-04 - acc: 0.8254 - val_loss: 0.0012 - val_acc: 0.7827
Epoch 7/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.6688e-04 - acc: 0.8480Epoch 00006: val_loss improved from 0.00115 to 0.00114, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.6819e-04 - acc: 0.8499 - val_loss: 0.0011 - val_acc: 0.7804
Epoch 8/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.5901e-04 - acc: 0.8389Epoch 00007: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.5909e-04 - acc: 0.8388 - val_loss: 0.0011 - val_acc: 0.7827
Epoch 9/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.7123e-04 - acc: 0.8347Epoch 00008: val_loss improved from 0.00114 to 0.00113, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.7073e-04 - acc: 0.8370 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 10/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.5915e-04 - acc: 0.8462Epoch 00009: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.6104e-04 - acc: 0.8440 - val_loss: 0.0011 - val_acc: 0.7897
Epoch 11/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4969e-04 - acc: 0.8468Epoch 00010: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4776e-04 - acc: 0.8475 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 12/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4447e-04 - acc: 0.8498Epoch 00011: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4556e-04 - acc: 0.8505 - val_loss: 0.0012 - val_acc: 0.7850
Epoch 13/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.4405e-04 - acc: 0.8522Epoch 00012: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4414e-04 - acc: 0.8528 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 14/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3998e-04 - acc: 0.8365Epoch 00013: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.4000e-04 - acc: 0.8382 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 15/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.3294e-04 - acc: 0.8486Epoch 00014: val_loss improved from 0.00113 to 0.00112, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.3249e-04 - acc: 0.8493 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 16/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0921e-04 - acc: 0.8582Epoch 00015: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0841e-04 - acc: 0.8575 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 17/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.1155e-04 - acc: 0.8654Epoch 00016: val_loss improved from 0.00112 to 0.00112, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 5.1349e-04 - acc: 0.8627 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 18/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0821e-04 - acc: 0.8522Epoch 00017: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.1194e-04 - acc: 0.8528 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 19/100
1664/1712 [============================>.] - ETA: 0s - loss: 5.0777e-04 - acc: 0.8522Epoch 00018: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 5.0609e-04 - acc: 0.8516 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 20/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9102e-04 - acc: 0.8480Epoch 00019: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9128e-04 - acc: 0.8499 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 21/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9032e-04 - acc: 0.8419Epoch 00020: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.9004e-04 - acc: 0.8429 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 22/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.9658e-04 - acc: 0.8504Epoch 00021: val_loss improved from 0.00112 to 0.00111, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.9553e-04 - acc: 0.8522 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 23/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8576e-04 - acc: 0.8582Epoch 00022: val_loss improved from 0.00111 to 0.00111, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.8390e-04 - acc: 0.8592 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 24/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8169e-04 - acc: 0.8612Epoch 00023: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8212e-04 - acc: 0.8604 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 25/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7433e-04 - acc: 0.8588Epoch 00024: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7269e-04 - acc: 0.8586 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 26/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7589e-04 - acc: 0.8540Epoch 00025: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7504e-04 - acc: 0.8569 - val_loss: 0.0011 - val_acc: 0.7944
Epoch 27/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8253e-04 - acc: 0.8516Epoch 00026: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8226e-04 - acc: 0.8528 - val_loss: 0.0012 - val_acc: 0.8037
Epoch 28/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7168e-04 - acc: 0.8696Epoch 00027: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7084e-04 - acc: 0.8662 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 29/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6889e-04 - acc: 0.8666Epoch 00028: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6940e-04 - acc: 0.8651 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 30/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.8483e-04 - acc: 0.8588Epoch 00029: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.8541e-04 - acc: 0.8604 - val_loss: 0.0012 - val_acc: 0.8061
Epoch 31/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.7074e-04 - acc: 0.8564Epoch 00030: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.7152e-04 - acc: 0.8569 - val_loss: 0.0011 - val_acc: 0.8131
Epoch 32/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5406e-04 - acc: 0.8630Epoch 00031: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5216e-04 - acc: 0.8651 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 33/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6803e-04 - acc: 0.8504Epoch 00032: val_loss improved from 0.00111 to 0.00110, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.6734e-04 - acc: 0.8511 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 34/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6518e-04 - acc: 0.8564Epoch 00033: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6474e-04 - acc: 0.8563 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 35/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.6490e-04 - acc: 0.8660Epoch 00034: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.6513e-04 - acc: 0.8662 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 36/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4759e-04 - acc: 0.8600Epoch 00035: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4954e-04 - acc: 0.8592 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 37/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4735e-04 - acc: 0.8642Epoch 00036: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4703e-04 - acc: 0.8639 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 38/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5897e-04 - acc: 0.8642Epoch 00037: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5947e-04 - acc: 0.8627 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 39/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5355e-04 - acc: 0.8684Epoch 00038: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5220e-04 - acc: 0.8668 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 40/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5156e-04 - acc: 0.8576Epoch 00039: val_loss improved from 0.00110 to 0.00110, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.5083e-04 - acc: 0.8569 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 41/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4755e-04 - acc: 0.8714Epoch 00040: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4719e-04 - acc: 0.8715 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 42/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.5332e-04 - acc: 0.8678Epoch 00041: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.5297e-04 - acc: 0.8674 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 43/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.4585e-04 - acc: 0.8648Epoch 00042: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.4711e-04 - acc: 0.8662 - val_loss: 0.0011 - val_acc: 0.8131
Epoch 44/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2895e-04 - acc: 0.8588Epoch 00043: val_loss improved from 0.00110 to 0.00109, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.2960e-04 - acc: 0.8598 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 45/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2244e-04 - acc: 0.8600Epoch 00044: val_loss improved from 0.00109 to 0.00109, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.2231e-04 - acc: 0.8563 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 46/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2690e-04 - acc: 0.8684Epoch 00045: val_loss improved from 0.00109 to 0.00109, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.2907e-04 - acc: 0.8709 - val_loss: 0.0011 - val_acc: 0.8154
Epoch 47/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2305e-04 - acc: 0.8642Epoch 00046: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2196e-04 - acc: 0.8668 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 48/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2568e-04 - acc: 0.8594Epoch 00047: val_loss improved from 0.00109 to 0.00108, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.3026e-04 - acc: 0.8598 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 49/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2655e-04 - acc: 0.8528Epoch 00048: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2526e-04 - acc: 0.8522 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 50/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2105e-04 - acc: 0.8684Epoch 00049: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2159e-04 - acc: 0.8651 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 51/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1841e-04 - acc: 0.8636Epoch 00050: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1837e-04 - acc: 0.8645 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 52/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1580e-04 - acc: 0.8642Epoch 00051: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1689e-04 - acc: 0.8639 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 53/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2107e-04 - acc: 0.8762Epoch 00052: val_loss improved from 0.00108 to 0.00107, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.2067e-04 - acc: 0.8773 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 54/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1646e-04 - acc: 0.8798Epoch 00053: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1746e-04 - acc: 0.8797 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 55/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1614e-04 - acc: 0.8720Epoch 00054: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1806e-04 - acc: 0.8715 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 56/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1842e-04 - acc: 0.8618Epoch 00055: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1890e-04 - acc: 0.8616 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 57/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.2227e-04 - acc: 0.8678Epoch 00056: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2193e-04 - acc: 0.8692 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 58/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1864e-04 - acc: 0.8702Epoch 00057: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1856e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 59/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1601e-04 - acc: 0.8636Epoch 00058: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1658e-04 - acc: 0.8633 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 60/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0995e-04 - acc: 0.8684Epoch 00059: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0885e-04 - acc: 0.8674 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 61/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0720e-04 - acc: 0.8828Epoch 00060: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0810e-04 - acc: 0.8826 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 62/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0760e-04 - acc: 0.8648Epoch 00061: val_loss improved from 0.00107 to 0.00107, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 4.0731e-04 - acc: 0.8639 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 63/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.3022e-04 - acc: 0.8726Epoch 00062: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.2854e-04 - acc: 0.8727 - val_loss: 0.0012 - val_acc: 0.8014
Epoch 64/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1838e-04 - acc: 0.8726Epoch 00063: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1657e-04 - acc: 0.8721 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 65/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0084e-04 - acc: 0.8732Epoch 00064: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0027e-04 - acc: 0.8738 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 66/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0767e-04 - acc: 0.8594Epoch 00065: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0667e-04 - acc: 0.8569 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 67/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9821e-04 - acc: 0.8780Epoch 00066: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9841e-04 - acc: 0.8768 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 68/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0911e-04 - acc: 0.8714Epoch 00067: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0822e-04 - acc: 0.8686 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 69/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9450e-04 - acc: 0.8762Epoch 00068: val_loss improved from 0.00107 to 0.00106, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 3.9388e-04 - acc: 0.8762 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 70/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0241e-04 - acc: 0.8696Epoch 00069: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0222e-04 - acc: 0.8697 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 71/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9630e-04 - acc: 0.8768Epoch 00070: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9518e-04 - acc: 0.8779 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 72/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9273e-04 - acc: 0.8780Epoch 00071: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9373e-04 - acc: 0.8797 - val_loss: 0.0011 - val_acc: 0.8107
Epoch 73/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9262e-04 - acc: 0.8798Epoch 00072: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9315e-04 - acc: 0.8803 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 74/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0031e-04 - acc: 0.8648Epoch 00073: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9894e-04 - acc: 0.8651 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 75/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0010e-04 - acc: 0.8720Epoch 00074: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9965e-04 - acc: 0.8715 - val_loss: 0.0011 - val_acc: 0.8131
Epoch 76/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.1185e-04 - acc: 0.8774Epoch 00075: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.1064e-04 - acc: 0.8785 - val_loss: 0.0011 - val_acc: 0.7991
Epoch 77/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9108e-04 - acc: 0.8648Epoch 00076: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9066e-04 - acc: 0.8657 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 78/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0191e-04 - acc: 0.8690Epoch 00077: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0355e-04 - acc: 0.8715 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 79/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9226e-04 - acc: 0.8618Epoch 00078: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9220e-04 - acc: 0.8604 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 80/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8314e-04 - acc: 0.8708Epoch 00079: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8247e-04 - acc: 0.8715 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 81/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.7934e-04 - acc: 0.8678Epoch 00080: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.7934e-04 - acc: 0.8668 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 82/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8237e-04 - acc: 0.8762Epoch 00081: val_loss improved from 0.00106 to 0.00105, saving model to my_model.adamax.h5
1712/1712 [==============================] - 0s - loss: 3.8322e-04 - acc: 0.8762 - val_loss: 0.0010 - val_acc: 0.8131
Epoch 83/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8821e-04 - acc: 0.8846Epoch 00082: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8849e-04 - acc: 0.8849 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 84/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0225e-04 - acc: 0.8774Epoch 00083: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 4.0114e-04 - acc: 0.8773 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 85/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9163e-04 - acc: 0.8732Epoch 00084: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8900e-04 - acc: 0.8744 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 86/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8478e-04 - acc: 0.8690Epoch 00085: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8654e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 87/100
1664/1712 [============================>.] - ETA: 0s - loss: 4.0165e-04 - acc: 0.8756Epoch 00086: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9986e-04 - acc: 0.8773 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 88/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9051e-04 - acc: 0.8744Epoch 00087: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8850e-04 - acc: 0.8721 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 89/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8655e-04 - acc: 0.8816Epoch 00088: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8697e-04 - acc: 0.8803 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 90/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8580e-04 - acc: 0.8810Epoch 00089: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8497e-04 - acc: 0.8826 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 91/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8697e-04 - acc: 0.8582Epoch 00090: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8633e-04 - acc: 0.8581 - val_loss: 0.0011 - val_acc: 0.8061
Epoch 92/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8190e-04 - acc: 0.8540Epoch 00091: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8217e-04 - acc: 0.8551 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 93/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8817e-04 - acc: 0.8678Epoch 00092: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8743e-04 - acc: 0.8674 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 94/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.7966e-04 - acc: 0.8660Epoch 00093: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.7965e-04 - acc: 0.8680 - val_loss: 0.0011 - val_acc: 0.7967
Epoch 95/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.9379e-04 - acc: 0.8822Epoch 00094: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.9296e-04 - acc: 0.8814 - val_loss: 0.0011 - val_acc: 0.8014
Epoch 96/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.7447e-04 - acc: 0.8714Epoch 00095: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.7554e-04 - acc: 0.8692 - val_loss: 0.0011 - val_acc: 0.8084
Epoch 97/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.7888e-04 - acc: 0.8702Epoch 00096: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.7878e-04 - acc: 0.8697 - val_loss: 0.0010 - val_acc: 0.8061
Epoch 98/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8240e-04 - acc: 0.8738Epoch 00097: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8266e-04 - acc: 0.8727 - val_loss: 0.0010 - val_acc: 0.7991
Epoch 99/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.8187e-04 - acc: 0.8684Epoch 00098: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8022e-04 - acc: 0.8697 - val_loss: 0.0011 - val_acc: 0.8037
Epoch 100/100
1664/1712 [============================>.] - ETA: 0s - loss: 3.7981e-04 - acc: 0.8864Epoch 00099: val_loss did not improve
1712/1712 [==============================] - 0s - loss: 3.8001e-04 - acc: 0.8832 - val_loss: 0.0010 - val_acc: 0.8061

Step 7: Visualize the Loss and Test Predictions

(IMPLEMENTATION) Answer a few questions and visualize the loss

Question 1: Outline the steps you took to get to your final neural network architecture and your reasoning at each step.

Answer: I will test the model with all optimizers of SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax, Nadam. Parameter values used for building the architecture are typical.

Step 1: The problem being solved is on computer vision, a CNN architecture seems to be a good fit for this problem. Design a simple CNN : (,96,96,1) ---> Conv2D (kernel size = 16) --> MaxPooling ( size = 2)
-->Flatten --> Dropout(0.3) ---> Dense ---> 30 .

         => min validation loss = 0.004, overfitting signal begins at epoch of 5

Step 2: Add one more layer, and increase Dropout value to reduce overfitting . Design a simple CNN : (,96,96,1) ---> Conv2D (kernel size = 32) --> MaxPooling ( size = 2) ---> Conv2D (kernel size = 12) --> MaxPooling ( size = 2) -->Flatten --> Dropout(0.5) ---> Dense ---> 30

         => min validation loss = 0.001, overfitting is not shown after 50 epoch

Question 2: Defend your choice of optimizer. Which optimizers did you test, and how did you determine which worked best?

Answer: As shown, adamax optimizer can reach min loss of 0.00104809582677 among all tested optimizers of SGD, RMSprop, Adagrad, Adadelta, Adam, Adamax after all. I decide to pick adamax optimizer to contribute to my training.

Use the code cell below to plot the training and validation loss of your neural network. You may find this resource useful.

In [20]:
## TODO: Visualize the training and validation loss of your neural network
import matplotlib.pyplot as plt

# summarize history for accuracy
for opt in opti:
    plt.plot(hist[opt].history['acc'])
    plt.plot(hist[opt].history['val_acc'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.savefig("Accuracy_With_Optimizers.png")
plt.legend(opti, loc='center left', bbox_to_anchor=(1, 0.5))
plt.show()
In [21]:
## TODO: Visualize the training and validation loss of your neural network
import matplotlib.pyplot as plt
opti =['rmsprop','adam','sgd','adagrad','adadelta','adamax']
# summarize history for accuracy
for opt in opti:
    plt.plot(hist[opt].history['val_acc'])
plt.title('model val accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.savefig("Accuracy_With_Optimizers.png")
plt.legend(opti, loc='center left', bbox_to_anchor=(1, 0.5))
plt.show()
In [22]:
for opt in opti:
    plt.plot(hist[opt].history['loss'])
    plt.plot(hist[opt].history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(opti, loc='center left', bbox_to_anchor=(1, 0.5))
plt.savefig("Loss_With_Optimizers.png")
plt.show()
In [23]:
for opt in opti:
    plt.plot(hist[opt].history['loss'])
    plt.plot(hist[opt].history['val_loss'])
plt.title('model  val loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.ylim(0.000, 0.005)
plt.legend(opti, loc='center left', bbox_to_anchor=(1, 0.5))
plt.savefig("Closer_Loss_With_Optimizers.png")
plt.show()
In [26]:
for opt in opti:
    plt.plot(hist[opt].history['val_loss'])
plt.title('model  val loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.ylim(0.000, 0.005)
plt.legend(opti, loc='center left', bbox_to_anchor=(1, 0.5))
plt.savefig("Validation_Loss_With_Optimizers.png")
plt.show()
In [16]:
min_val_loss = []
for i in hist:
    min_val_loss.append(min(hist[i].history['val_loss']))
    print(i,min(hist[i].history['val_loss']))
print(min(min_val_loss))

sgd 0.0011002618393
adam 0.00109554069643
adagrad 0.00119990664186
rmsprop 0.00136565934501
adamax 0.00104809582677
adadelta 0.00117059798572
0.00104809582677

Training and Test Loss Graph for CNN with optimizer Adamax

In [32]:
opti =['rmsprop','adam','sgd','adagrad','adadelta','adamax']
opt = opti[5]
plt.plot(hist[opt].history['val_loss'])
plt.plot(hist[opt].history['loss'])
plt.title('model  val loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.ylim(0.000, 0.005)
plt.legend(['test','train'], loc='center left', bbox_to_anchor=(1, 0.5))
plt.savefig("Validation_Loss_With_Optimizers.png")
plt.show()
Question 3: Do you notice any evidence of overfitting or underfitting in the above plot? If so, what steps have you taken to improve your model? Note that slight overfitting or underfitting will not hurt your chances of a successful submission, as long as you have attempted some solutions towards improving your model (such as regularization, dropout, increased/decreased number of layers, etc).

Answer: There seems to be overfitting , around epoch 5. So I added one more Conv2D layer, and increase drop out from 0.3 to 0.5. I seem to get overfitting around epoch 10 . Test loss tends to go parallel or even divergent with train loss, rather than merge to 0. I trained this with early stopping with patience of 30 epochs , so it actually has some improvements at least until epoch of 70.

Visualize a Subset of the Test Predictions

Execute the code cell below to visualize your model's predicted keypoints on a subset of the testing images.

In [66]:
opti =['rmsprop','adam','sgd','adagrad','adadelta','adamax']
model.load_weights('my_model.{}.h5'.format(opti[5]))
In [67]:
y_test = model.predict(X_test)
fig = plt.figure(figsize=(20,20))
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
move = 100
for i in range(9):
    ax = fig.add_subplot(3, 3, i + 1, xticks=[], yticks=[])
    plot_data(X_test[move+i], y_test[move+i], ax)

Step 8: Complete the pipeline

With the work you did in Sections 1 and 2 of this notebook, along with your freshly trained facial keypoint detector, you can now complete the full pipeline. That is given a color image containing a person or persons you can now

  • Detect the faces in this image automatically using OpenCV
  • Predict the facial keypoints in each face detected in the image
  • Paint predicted keypoints on each face detected

In this Subsection you will do just this!

(IMPLEMENTATION) Facial Keypoints Detector

Use the OpenCV face detection functionality you built in previous Sections to expand the functionality of your keypoints detector to color images with arbitrary size. Your function should perform the following steps

  1. Accept a color image.
  2. Convert the image to grayscale.
  3. Detect and crop the face contained in the image.
  4. Locate the facial keypoints in the cropped image.
  5. Overlay the facial keypoints in the original (color, uncropped) image.

Note: step 4 can be the trickiest because remember your convolutional network is only trained to detect facial keypoints in $96 \times 96$ grayscale images where each pixel was normalized to lie in the interval $[0,1]$, and remember that each facial keypoint was normalized during training to the interval $[-1,1]$. This means - practically speaking - to paint detected keypoints onto a test face you need to perform this same pre-processing to your candidate face - that is after detecting it you should resize it to $96 \times 96$ and normalize its values before feeding it into your facial keypoint detector. To be shown correctly on the original image the output keypoints from your detector then need to be shifted and re-normalized from the interval $[-1,1]$ to the width and height of your detected face.

When complete you should be able to produce example images like the one below

In [68]:
# Import required libraries for this section

%matplotlib inline

import numpy as np
import matplotlib.pyplot as plt
import math
import cv2                     # OpenCV library for computer vision
from PIL import Image
import time
In [69]:
# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')
image_copy = np.copy(image)
# Convert the image to RGB colorspace
rgb = cv2.cvtColor(image_copy, cv2.COLOR_BGR2RGB)
# plot our image
fig = plt.figure(figsize = (9,9))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('image copy')
ax1.imshow(rgb)
Out[69]:
<matplotlib.image.AxesImage at 0x7fd6c7ece160>

Useful Functions for Following Assignment

In [70]:
def face_detector(image,if_gray=False,face_detection_val = (1.26,6)):
    ## TODO: Run the face detector on the de-noised image to improve your detections and display the result
    # Convert the RGB  image to grayscale
    if if_gray == True:
        gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
    else:
        gray = np.copy(image)
    # Extract the pre-trained face detector from an xml file
    face_cascade = cv2.CascadeClassifier('detector_architectures/haarcascade_frontalface_default.xml')

    # Detect the faces in image
    
    faces = face_cascade.detectMultiScale(gray, face_detection_val[0], face_detection_val[1])

    # Print the number of faces detected in the image
    print('Number of faces detected:', len(faces))
    return faces

def draw_faces(image,faces):
    image_with_faces = np.copy(image)

    # Get the bounding box for each detected face
    for (x,y,w,h) in faces:
        # Add a red bounding box to the detections image
        cv2.rectangle(image_with_faces, (x,y), (x+w,y+h), (255,0,0), 3)
        #print(x,y,w,h)
    return image_with_faces

def face_extractor(image,faces):
    image_with_faces = np.copy(image)
    the_faces = []

    # Get the bounding box for each detected face
    for (x,y,w,h) in faces:
        # Add a red bounding box to the detections image
        
        the_faces.append(image_with_faces[y:y+h,x:x+w])
        #print(x,y,w,h)
    return the_faces

def make_image_gray(image,if_bgr =False):
    img = np.copy(image)
    rgb = None
    if if_bgr:
        rgb = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    else:
        rgb = np.copy(img)
     
    gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
    return gray
    
def normalize_gray_val(image):
    img = np.copy(image)
    img = img/255
    return img

def reshape_image(image,shape = (96,96)):
    img = np.copy(image)
    img = cv2.resize(img, shape, interpolation = cv2.INTER_CUBIC) 
    new_img = np.ndarray(shape=(1,shape[0],shape[1],1), dtype=float, order='F')
    new_img[0,:,:,0] = img
    return new_img
# Prepare gray 96x96 [0,1] face images from an original image
def face_data_prepare_96x96(image,if_bgr = True):
    img = np.copy(image)
    
    if if_bgr:
        rgb = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    else:
        rgb =img
    faces = face_detector(rgb)
    img_with_faces = draw_faces(rgb,faces)
    the_faces = face_extractor(rgb,faces)
    
    for i,face in enumerate(the_faces):
        the_faces[i] = make_image_gray(face)
        the_faces[i] = normalize_gray_val(the_faces[i])
        the_faces[i] = reshape_image(the_faces[i])
        
    return the_faces,faces,image_with_faces
# Draw dots on gray 96x96 [0,1] face images 
def face_dot_drawing_96x96(the_faces,figsize = (16,16)):
    fig = plt.figure(figsize=figsize)
    for i,face in enumerate(the_faces):
        y_test = model.predict(face)[0]
        ax = fig.add_subplot(2,2,i+1, xticks=[], yticks=[])
        plot_data(face, y_test, ax)
        
# Return key point coordinate in original image
def return_key_points(image,if_bgr = True):
    face_images,face_coords,image_with_faces = face_data_prepare_96x96(image,if_bgr)
    faces_key_points = []
    
    for i,face_image in enumerate(face_images):
        distorted_key_points = model.predict(face_image)[0]
        (x,y,w,h) = face_coords[i]
        x_set = distorted_key_points[0::2] * (w / 2) + (w / 2) + x
        y_set = distorted_key_points[1::2] * (h / 2) + (h / 2) + y
        faces_key_points.append((x_set,y_set)) 
    foo = np.array(faces_key_points)
    print('num key points',len(faces_key_points))
    print(foo.shape)
    return faces_key_points,face_coords,image_with_faces

def draw_dots_on_faces_in_original_image(image,faces_key_points,if_bgr=True):
    img = np.copy(image)
    if if_bgr:
        rgb = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    else:
        rgb = img
   
    fig = plt.figure(figsize=(8,8))
    ax1 = fig.add_subplot(111)
    #ax1.set_xticks([])
    #ax1.set_yticks([])
    for i,(x_set,y_set) in enumerate(faces_key_points):
        print('xxxxx',x_set.shape)
        ax1.scatter(x_set[:],y_set[:],marker='o', c='c', s=4)
    ax1.imshow(rgb)
    
def draw_dots_on_faces_in_video_frame(frame,faces_key_points,if_bgr=True,):
    fr = np.copy(frame)
    if if_bgr:
        rgb = cv2.cvtColor(fr,cv2.COLOR_BGR2RGB)
    else:
        rgb = fr
    for (x_set,y_set) in faces_key_points:
        for (x,y) in zip(x_set,y_set):
            cv2.circle(fr,(x,y), 3, (0,0,255), -1)
    return fr
In [71]:
### TODO: Use the face detection code we saw in Section 1 with your trained conv-net 
## TODO : Paint the predicted keypoints on the test image
image = cv2.imread('huy.png')
image = cv2.imread('images/obamas4.jpg')
the_faces,_,_ = face_data_prepare_96x96(image)
face_dot_drawing_96x96(the_faces)

Number of faces detected: 2
In [72]:
faces_key_points,_,_ = return_key_points(image)
draw_dots_on_faces_in_original_image(image,faces_key_points)

Number of faces detected: 2
num key points 2
(2, 2, 15)
xxxxx (15,)
xxxxx (15,)

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera

Now you can add facial keypoint detection to your laptop camera - as illustrated in the gif below.

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for keypoint detection and marking in the previous exercise and you should be good to go!

In [76]:
import cv2
import time 
from keras.models import load_model
def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face dotting activated")
    vc = cv2.VideoCapture(0)

    # Try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # keep video stream open
    while rval:
        # plot image from camera with detections marked
        faces_key_points,_,frame_with_face = return_key_points(frame)
        #frame = draw_dots_on_faces_in_video_frame(frame,faces_key_points)
        frame = draw_dots_on_faces_in_video_frame(frame,faces_key_points)
        #print(frame.shape)
        cv2.imshow("face dotting activated", frame)
        
        # exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key == 27: # exit by pressing esc
            # destroy windows
            cv2.destroyAllWindows()
            
            # hack from stack overflow for making sure window closes on osx --> https://stackoverflow.com/questions/6116564/destroywindow-does-not-close-window-on-mac-using-python-and-opencv
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()
In [77]:
# Run your keypoint face painter
#laptop_camera_go()

Face with Dots

(Optional) Further Directions - add a filter using facial keypoints

Using your freshly minted facial keypoint detector pipeline you can now do things like add fun filters to a person's face automatically. In this optional exercise you can play around with adding sunglasses automatically to each individual's face in an image as shown in a demonstration image below.

To produce this effect an image of a pair of sunglasses shown in the Python cell below.

In [80]:
# Load in sunglasses image - note the usage of the special option
# cv2.IMREAD_UNCHANGED, this option is used because the sunglasses 
# image has a 4th channel that allows us to control how transparent each pixel in the image is
sunglasses = cv2.imread("images/sunglasses_4.png", cv2.IMREAD_UNCHANGED)

# Plot the image
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
#ax1.set_xticks([])
#ax1.set_yticks([])
ax1.imshow(sunglasses)
#ax1.axis('off');
Out[80]:
<matplotlib.image.AxesImage at 0x7fd6c7da0898>

This image is placed over each individual's face using the detected eye points to determine the location of the sunglasses, and eyebrow points to determine the size that the sunglasses should be for each person (one could also use the nose point to determine this).

Notice that this image actually has 4 channels, not just 3.

In [81]:
# Print out the shape of the sunglasses image
print ('The sunglasses image has shape: ' + str(np.shape(sunglasses)))

The sunglasses image has shape: (1123, 3064, 4)

It has the usual red, blue, and green channels any color image has, with the 4th channel representing the transparency level of each pixel in the image. Here's how the transparency channel works: the lower the value, the more transparent the pixel will become. The lower bound (completely transparent) is zero here, so any pixels set to 0 will not be seen.

This is how we can place this image of sunglasses on someone's face and still see the area around of their face where the sunglasses lie - because these pixels in the sunglasses image have been made completely transparent.

Lets check out the alpha channel of our sunglasses image in the next Python cell. Note because many of the pixels near the boundary are transparent we'll need to explicitly print out non-zero values if we want to see them.

In [82]:
# Print out the sunglasses transparency (alpha) channel
alpha_channel = sunglasses[:,:,3]
print ('the alpha channel here looks like')
print (alpha_channel)

# Just to double check that there are indeed non-zero values
# Let's find and print out every value greater than zero
values = np.where(alpha_channel != 0)
print ('\n the non-zero values of the alpha channel look like')
print (values)

the alpha channel here looks like
[[0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]]

 the non-zero values of the alpha channel look like
(array([  17,   17,   17, ..., 1109, 1109, 1109]), array([ 687,  688,  689, ..., 2376, 2377, 2378]))

This means that when we place this sunglasses image on top of another image, we can use the transparency channel as a filter to tell us which pixels to overlay on a new image (only the non-transparent ones with values greater than zero).

One last thing: it's helpful to understand which keypoint belongs to the eyes, mouth, etc. So, in the image below, we also display the index of each facial keypoint directly on the image so that you can tell which keypoints are for the eyes, eyebrows, etc.

With this information, you're well on your way to completing this filtering task! See if you can place the sunglasses automatically on the individuals in the image loaded in / shown in the next Python cell.

In [87]:
# Load in color image for face detection
image = cv2.imread('images/obamas4.jpg')

# Convert the image to RGB colorspace
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)


# Plot the image
fig = plt.figure(figsize = (8,8))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Original Image')
ax1.imshow(image)
Out[87]:
<matplotlib.image.AxesImage at 0x7fd6c7eb1208>
In [88]:
def transform_glass2eyes(image,sunglasses_image,faces_key_points):
    
    img = np.copy(image)
    sunglasses = np.copy(sunglasses_image)
    eye_key_points = []
    print(np.array(eye_key_points).shape)
    # Return the vertices of the bounding 
    for x_set,y_set in faces_key_points:
        eye_key_points = []
        for (x_point,y_point) in zip(x_set[0:10], y_set[0:10]):
                eye_key_points.append((x_point, y_point))
        eye_bounding_rect = cv2.boundingRect(np.array(eye_key_points).astype(np.float32))
        glasses_triangle_vertices = np.array([(260,220),(2820,220),(260,750)]).astype(np.float32)
        eyes_triangle_vertices = np.array([(eye_bounding_rect[0],eye_bounding_rect[1]),
                                           (eye_bounding_rect[0]+eye_bounding_rect[2],eye_bounding_rect[1]),
                                           (eye_bounding_rect[0],eye_bounding_rect[1]+eye_bounding_rect[3])]).astype(np.float32)
        # Compute Affine Transform matrix T
        T= cv2.getAffineTransform(glasses_triangle_vertices, eyes_triangle_vertices)
        # Apply Affine Transform matrix T to warp the sunglasses image, adapt to size of the image with faces
        transformed_sunglasses = cv2.warpAffine(sunglasses, T , (img.shape[1], img.shape[0]))
        # Pick none-zero valued entries in the sunglasses image
        transformed_sunglasses_mask = transformed_sunglasses[:,:,3] > 0
        # Replace the image with faces with the transformed sunglasses
        img[:,:,:][transformed_sunglasses_mask] = transformed_sunglasses[:,:,0:3][transformed_sunglasses_mask]
    return img
In [90]:
## (Optional) TODO: Use the face detection code we saw in Section 1 with your trained conv-net to put
## sunglasses on the individuals in our test image
faces_key_points,_,_ = return_key_points(image)
#draw_dots_on_faces_in_original_image(image,faces_key_points)
image = transform_glass2eyes(image,sunglasses,faces_key_points)
fig = plt.figure(figsize = (6,6))
ax1 = fig.add_subplot(111)
ax1.set_xticks([])
ax1.set_yticks([])
ax1.set_title('Face and Sunglasses')
ax1.imshow(img)

Number of faces detected: 2
num key points 2
(2, 2, 15)
(0,)
Out[90]:
<matplotlib.image.AxesImage at 0x7fd6c7e6a1d0>

(Optional) Further Directions - add a filter using facial keypoints to your laptop camera

Now you can add the sunglasses filter to your laptop camera - as illustrated in the gif below.

The next Python cell contains the basic laptop video camera function used in the previous optional video exercises. Combine it with the functionality you developed for adding sunglasses to someone's face in the previous optional exercise and you should be good to go!

In [ ]:
import cv2
import time 
from keras.models import load_model
import numpy as np

def laptop_camera_go():
    # Create instance of video capturer
    cv2.namedWindow("face detection activated")
    vc = cv2.VideoCapture(0)

    # try to get the first frame
    if vc.isOpened(): 
        rval, frame = vc.read()
    else:
        rval = False
    
    # Keep video stream open
    while rval:
        # Plot image from camera with detections marked
        faces_key_points,_,_ = return_key_points(frame)
        draw_dots_on_faces_in_original_image(frame,faces_key_points)
        frame = transform_glass2eyes(frame,sunglasses,faces_key_points)
        cv2.imshow("face detection activated", frame)
        
        # Exit functionality - press any key to exit laptop video
        key = cv2.waitKey(20)
        if key == 27: # exit by pressing key esc
            # Destroy windows 
            cv2.destroyAllWindows()
            
            for i in range (1,5):
                cv2.waitKey(1)
            return
        
        # Read next frame
        time.sleep(0.05)             # control framerate for computation - default 20 frames per sec
        rval, frame = vc.read()
In [ ]:
# Load facial landmark detector model
model = load_model('my_model.adamax.h5')

# Run sunglasses painter
laptop_camera_go()

Face with Sunglasses